0be a scalar function conditioned only on c. Consider the loss function:\\\\nLNCE\\\\nθ(x,c) =−Ep(x,c)logσ(rθ(x,c))−λcEp(x)p(c)logσ(−rθ(x,c)). (15)\\\\nGiven unlimited model expressivity for rθ, the optimal solution for minimizing LNCE\\\\nθsatisfies\\\\nr∗\\\\nθ(x,c) = logp(x|c)\\\\np(x)−logλc. (16)\\\\nProof. We omit the full proof here, as it requires only a redefinition of the distributions Qx,cfrom\\\\nthe proof of Theorem A.1:\\\\nQx,c:={p(x,c)\\\\np(x,c) +λcp(x)p(c),λcp(x)p(c)\\\\np(x,c) +λcp(x)p(c)}={p(x|c)\\\\np(x|c) +λcp(x),λcp(x)\\\\np(x|c) +λcp(x)}\\\\nThen we can follow the steps in the proof of Theorem A.1 to arrive at the result.\\\\nIf let λc:=Z(c)1\\\\ns=\\\\x02\\\\nEp(x|c)[p(x|c)\\\\np(x)]s\\\\x031\\\\ns, we could guarantee the convergence of psample\\\\nθ topsample.\\\\nHowever, in practice estimating Z(c)could be intricately difficult, so we formalize λcas a training\\\\nparameter, resulting in our practical algorithm in Eq. 12.\\\\nC A DDITIONAL EXPERIMENTAL RESULTS\\\\nWe provide more image samples to compare CCA and CFG in Figure 7 and Figure 8.\\\\nWe illustrate the effect of training parameter βon the FID-IS trade-off in Figure 9. Overall, βaffects\\\\nthe fidelity-diversity trade-off similar to CCA λand the CFG method.\\\\n/uni00000018/uni00000013 /uni00000014/uni00000013/uni00000013 /uni00000014/uni00000018/uni00000013 /uni00000015/uni00000013/uni00000013 /uni00000015/uni00000018/uni00000013 /uni00000016/uni00000013/uni00000013 /uni00000016/uni00000018/uni00000013 /uni00000017/uni00000013/uni00000013\\\\n/uni0000002c/uni00000036/uni00000018/uni00000014/uni00000013/uni00000014/uni00000018/uni00000029/uni0000002c/uni00000027\\\\n/uni00000030/uni00000052/uni00000047/uni00000048/uni0000004f\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni0000000b/uni0000004"}}},{"rowIdx":74,"cells":{"Researcher Name":{"kind":"string","value":"Sahil Singla"},"ORCID":{"kind":"string","value":"0000-0002-8800-6479"},"Researcher Query Keyword":{"kind":"string","value":"Optimal Pricing with Limited Sampling"},"Research Full Paper":{"kind":"string","value":"{'Online Combinatorial Allocations and Auctions with Few Samples': 'Title: Online Combinatorial Allocations and Auctions with Few Samples\\\\nAbstract\\\\nWe consider the online problem in which an intermediary trades identical items with a\\\\nsequence of n buyers and n sellers, each of unit demand. We assume that the values of\\\\nthe traders are selected by an adversary and the sequence is randomly permuted. We give\\\\ncompetitive algorithms for two objectives: welfare and gain-from-trade.\\\\n1 \\\\nIntroduction\\\\nWe study the problem of facilitating trade between n buyers and n sellers that arrive online.\\\\nWe consider one of the simplest settings in which each trader, buyer or seller, is interested in\\\\ntrading a single item, and all items are identical. Each trader has a value for the item; a seller\\\\nwill sell to any price higher than its value and a buyer will buy for any price lower than its\\\\nvalue. Upon encountering a trader, the online algorithm makes an irrevocable price offer, the\\\\ntrader reveals its value and, if the value is at the correct side of the offered price the item is\\\\ntraded. After buying an item from a seller, the online algorithm can store it indefinitely to sell\\\\nit to later buyers. Of course, the online algorithm can only sell to a buyer if it has at least one\\\\nitem at the time of the encounter.\\\\nWe consider online algorithms that offer prices based on the sequence of past values and\\\\nwe assume that the online algorithm knows only the number of buyers and sellers, but not\\\\ntheir values. The values of the sellers and buyers are selected adversarially and are randomly\\\\npermuted. In that respect, the problem is a generalization of the well-known secretary problem.\\\\nThe secretary problem corresponds to the special case in which there are only buyers, the\\\\nalgorithm starts with a single item, and the objective is to maximize the total welfare, which is\\\\nto give the value to a buyer with as high value as possible.\\\\nExtending this to both sellers and buyers, creates a substantially richer setting. One of the\\\\nmost important differences between the two settings is that besides the objective of maximizing\\\\nthe total welfare, we now have the objective of maximizing the gain-from-trade. For both\\\\nobjectives, the algorithm must buy from sellers with low values and sell to buyers with high\\\\nvalues. The objective is that at the end, the items end up at the hands of the traders, sellers or\\\\nbuyers, with the highest values. The welfare of a solution is defined as the value of the buyers\\\\nand sellers that have an item. The gain-from-trade of a solution is the difference between the\\\\nwelfare at the end of the process minus the welfare at the beginning. At optimality the two\\\\nobjectives are interchangeable: an algorithm achieves the maximum welfare if and only if it\\\\n∗Supported by the ERC Advanced Grant 321171 (ALGAME).\\\\n1\\\\nar\\\\nX\\\\niv\\\\n:1\\\\n81\\\\n2.\\\\n11\\\\n14\\\\n9v\\\\n1 \\\\n [c\\\\ns.G\\\\nT]\\\\n 2\\\\n8 D\\\\nec\\\\n 20\\\\n18\\\\nachieves the maximum gain-from-trade. But for approximate solutions, the two objectives are\\\\nentirely different, with the gain-from-trade being the most demanding one.\\\\nThe Bayesian version of the problem, in which the values of the buyers and sellers are\\\\ndrawn from known probability distributions has been extensively considered in the literature.\\\\nOptimal mechanisms for bilateral trading, that is, the offline case of a single seller and a single\\\\nbuyer, were first analysed by Myerson and Satterthwaite in [17] and played a pivotal role in the\\\\ndevelopment of the area. The online Bayesian case was considered in [9], where the values are\\\\ndrawn from a known distribution but the sequence is adversarially ordered.\\\\nA generalization of our model is when the items are not identical and each buyer has different\\\\nvalue for each one of them, i.e., each seller has a value for its item and each buyer has a vector\\\\nof values, one for every pair buyer-seller. This is also a generalization of the well-studied online\\\\nmaximum-matching problem [14, 12]. One can cast the online maximum-matching problem as\\\\nthe version in which the sellers arrive first and have zero value for their item. The optimal online\\\\nalgorithm for this problem has competitive ratio 1/e, when the objective is the welfare (which\\\\nin the absense of seller values is identical to the gain-from-trade). Our model is incomparable to\\\\nthe online maximum-matching problem: it is simpler in the sense that the items are identical (a\\\\nsingle value for each buyer instead of a vector of buyer-item values), and at the same time more\\\\ncomplicated in that the items are not present throughout the process, but they are brought to\\\\nthe market by sellers that have their own utility. The fact that in our model the buyer-item\\\\nvalues are related, allows for a much better competitive ratio regarding the welfare, (almost) 1\\\\ninstead of 1/e. More importantly, our algorithm is truthful, while in contrast, no good truthful\\\\nalgorithm is known for the online maximum-matching problem, which remains one of the main\\\\nopen problems of the area. On the other hand, the \\\\nintroduction of sellers poses new challenges,\\\\nespecially with respect to the objective of the gain-from-trade.\\\\nThere are also similarities between our model and the extension of the classical secretary\\\\nproblem to k secretaries. From an influential result by Kleinberg [13] we know that this problem\\\\nhas competitive ratio 1 − 1/√k which is asymptotically tight, and can be transformed into a\\\\ntruthful algorithm. This result depends strongly on the knowledge of k. In our case the equiva-\\\\nlent measure, the number of trades is not known from the beginning and has to be learned, with\\\\na degree of precision that is crucial, especially for the gain-from-trade objective. The fact that\\\\nthe gain-from-trade is not monotone as a function of time highlights the qualitative difference\\\\nbetween the two models; the gain-from-trade temporarily decreases when the algorithm buys\\\\nan item, with the risk of having a negative gain at the end. More generally, with the mix of\\\\nbuyers and sellers, wrong decisions are penalized more harshly and the monotone structure of\\\\nthe problem is disrupted.\\\\n1.1 Our \\\\nresults\\\\nWe consider the case when both the number of buyers and the number of sellers is n. For\\\\nwelfare we show a competitive ratio of 1− O˜(n−1/3), where O˜ hides logarithmic factors.\\\\nActually we can compare an online algorithm with two offline benchmarks: the optimal\\\\nbenchmark, in which all trades between buyers and sellers are possible, independently of their\\\\norder of appearance, and the expected sequential optimal in which an item can be transferred\\\\nfrom a seller to a buyer only if the seller precedes the buyer in the order.\\\\nOur online algorithm achieves a competitive ratio of 1−O˜(n−1/3) against the optimal bench-\\\\nmark. To achieve this, it has a small sampling phase of length O˜(n2/3) to estimate the median\\\\nof the values of all traders, and then uses it as a price for the remaining traders. But if the\\\\noptimal number of trades is small, such a scheme will fail to achieve competitive ratio almost\\\\none, because with constant probability there will not have enough items to sell to buyers with\\\\n2\\\\nhigh value. To deal with this risk, the algorithm not only samples values at the beginning but it\\\\nadditionally buys sufficiently many items, O˜(n2/3), from the first sellers1. The number O˜(n2/3)\\\\nof bought items balances the potential loss of the welfare that \\\\nresults from removing items from\\\\nsellers to the expected loss from not having enough items for buyers of high values.\\\\nThe term O(n−1/3) in the competitive ratio seems to be optimal for a scheme that fixes\\\\nthe price after the sampling phase and relates to the number of items needed to approximate\\\\nthe median to a good degree. It may be possible to improve this term to O(n−1/2) by a more\\\\nadaptive scheme, as in the case of the k-secretary problem [13]. Finally, it may be possible to\\\\nremove the logarithmic factors from the competitive ratio, but we have opted for simplicity and\\\\ncompleteness.\\\\nFor the objective of gain-from-trade, we give a truthful algorithm that has a constant com-\\\\npetitive ratio, assuming that the algorithm starts with an item. The competitive ratio is high,\\\\napproximately 103, but it drops to a small constant when the optimal number of trades is suffi-\\\\nciently high. The additional assumption of starting with an item is necessary, because without\\\\nit, no online algorithm can achieve a bounded competitive ratio.\\\\nThe main difficulty of designing an online algorithm for gain-from-trade is that even a single\\\\nitem that is left unsold at the end has dramatic effects on the gain-from-trade. The online\\\\nalgorithm must deal with the case of many traders, large welfare, but few optimal trades and\\\\nsmall gain-from-trade.\\\\nTo address this problem, our algorithm, unlike the case of welfare, has a large sampling\\\\nphase. It uses this phase to estimate the number of optimal trades and two prices for trading\\\\nwith buyers and sellers. If the expected number of optimal trades is high, the algorithm uses\\\\nthe two prices for trading with the remaining traders. But if the number is small, it runs the\\\\nsecretary algorithm with the item that it starts with.\\\\nThe analysis needs high concentration bounds on the expected number of trades to minimize\\\\nthe risk of having items left unsold. Our algorithm is ordinal, in the sense that it uses only the\\\\norder statistics of the values not the actual values themselves. This leaves little space for errors\\\\nand it may be possible that cardinal algorithms that use the actual values can do substantially\\\\nbetter.\\\\n1.2 Related Work\\\\nThe bilateral trade literature was initiated by Myerson and Satterthwaite in their seminal paper\\\\n[17]. They investigated the case of a single seller-buyer pair and proved their famous impos-\\\\nsibility result: there exists no truthful, individually rational and budget balanced mechanism\\\\nthat also maximizes the welfare (and consequently, the gain from trade). Subsequent research\\\\nstudied how well these two objectives can be approximated by relaxing these conditions. Blum-\\\\nrosen and Mizrahi [2] devised a 1/e-approximate, Bayesian incentive compatible mechanism\\\\nfor the gain from trade assuming the buyer’s valuation is monotone hazard rate. Brustle et\\\\nal. expanded in this direction in [3] for arbitrary valuations and downwards closed feasibility\\\\nconstraints over the allocations. In the case where there are multiple, unit demand, buyers\\\\nand sellers, McAfee provided a weakly budget balanced, 1 − 1/k approximate mechanism for\\\\nthe gain from trade in [15], where k is the number of trades in the optimal allocation. This\\\\nwas later extended to be strongly budget balanced by Segal-Halevi et al. in [18]. McAfee also\\\\nproved a simple 2-approximation to the gain from trade if the buyer’s median valuation is above\\\\n1Buying from the first sellers cannot be done truthfully unless the algorithm knows an upper bound on their\\\\nvalue. But this is not necessary since there is an alternative that has minor effects on the competitive ratio: the\\\\nalgorithm offers each seller the maximum value of the sellers so far. This is a truthful scheme that buys from all\\\\nbut a logarithmic number of sellers, in expectation.\\\\n3\\\\nthe seller’s [16]. This was significantly improved by Colini-Baldeschi et al. in [?] to 1/r and\\\\nO(log(1/r)), where r is the probability that the buyer’s valuation for the item is higher than\\\\nthe seller’s. Recently, Giannakopoulos et al. [9] studied an online variant of this setting where\\\\nbuyers and sellers are revealed sequentially by an adversary and have known prior distributions\\\\non the value of the items.\\\\nThe random order model we are using has its origins in the well-known secretary problem,\\\\nwhere n items arrive in online fashion and our goal is to maximize the probability of selecting\\\\nthe most valuable, without knowing their values in advance. The matroid secretary problem\\\\nwas introduced by Babaioff et al. [1]. In this setting, we are allowed to select more than item,\\\\nprovided our final selection satisfies matroid constraints. A variety of different matroids have\\\\nbeen studied, with many recent \\\\nresults presented by Dinitz in [6]. Of particular interest to\\\\nour problem are secretary problems on bipartite graphs. Here, the left hand side vertices of\\\\nthe graph are fixed and the right hand side vertices (along with their incident) edges appear\\\\nonline. The selected edges must form a (incomplete) matching and the goal is to maximize\\\\nthe sum of their weights. Babaioff et al. in [1] provided a 4d-competitive algorithm for the\\\\ntransversal matroid with bounded left degree d, which is a special case of the online bipartite\\\\nmatching where all edges connected to the same left hand side vertex have equal value. This\\\\nwas later improved to 16 by Dimitrov and Plaxton [5]. The case where all edges have unrelated\\\\nweights was first considered by Korula and Pal in [14] who designed a 8-competitive algorithm,\\\\nwhich was later improved to the optimal 1/e by Kesselheim et al. [12]. Another secretary\\\\nvariant which is close to our work is when the online selects k items instead of one, where\\\\nKleinberg [13] showed an asymptotically tight algorithm with competitive ratio 1−O(√1/k).\\\\nThe wide range of applications of secretary models (and the related prophet inequalities)\\\\nhave led to the design of posted price mechanisms, that are simple to describe, robust, truthful\\\\nand achieve surprisingly good approximation ratios. Hajiaghayi et al. introduced prophet\\\\ninequality techniques in online auction in [11]. The k-choice secretary described above was then\\\\nstudied in [10] which combined with [13] yielded an asympotically optimal, truthful mechanism.\\\\nFor more general auction settings, posted-price mechanisms have been used by Chawla et al.\\\\nin [4] for unit demand agents and expanded by Feldman et al. in [8] for combinatorial auctions\\\\nand [7] for online budgeted settings.\\\\n2 Model and Notation\\\\nThe setting of the random intermediation problem consists of sets B = {b1, . . . , bn} and S =\\\\n{s1, . . . , sn} containing the valuations of the buyers and sellers. For convenience, we assume\\\\nthat they are all distinct. The intermediary interacts with a uniformly random permutation σ\\\\nof B ∪S which is presented to him one agent at a time, over 2n steps. The intermediary has no\\\\nknowledge of σ(t) before step t. We use bi and sj to denote the i-th highest valued seller and\\\\nj-th lowest valued seller respectively.\\\\nWe study posted price mechanisms that upon seeing the identity of agent t offer price pt. This\\\\nprice can not depend on the entire valuation function; only the values within σ(1) . . . σ(t − 1)\\\\nwhich are revealed at this point. We buy or sell one item from sellers or buyers who accept our\\\\nprice, respectively. Of course, we can only sell items if we have stock available. Formally, let κt\\\\nbe the number of items at time t. Starting with κ0 items (with κ0 = 0 for welfare and 1 for the\\\\ngain-from-trade):\\\\nκt+1 =\\\\n\\\\uf8f1\\\\uf8f4\\\\uf8f4\\\\uf8f2\\\\uf8f4\\\\uf8f4\\\\uf8f3\\\\nκt + 1, if σ(t) ∈ S ∧ σ(t) ≤ pt\\\\nκt − 1, if σ(t) ∈ B ∧ κt ≥ 1 ∧ σ(t) ≥ pt\\\\nκt otherwise.\\\\n4\\\\nThe set of sellers from whom we bought items during the algorithm’s execution is\\\\nTS = {s ∈ S | ∃t σ(t) = s ≤ pt }\\\\nand similarly the set of buyers we sold to is TB = {b ∈ B | ∃t σ(t) = b ≥ pt ∧ κt > 0}. Notice\\\\nthat these are random variables, depending on σ.\\\\nThe social welfare of online algorithm A is the sum of the valuations of all agents with\\\\nitems. In particular, after executing A it is: WA(S,B) = E\\\\n[∑\\\\ns∈S\\\\\\\\TS s+\\\\n∑\\\\nb∈TB b\\\\n]\\\\n. The gain\\\\nfrom trade (or GFT) produced by algorithm A throughout the run is the difference between\\\\nthe final and starting welfare: GFTA(S,B) = E\\\\n[∑\\\\nb∈TB b−\\\\n∑\\\\ns∈TS s\\\\n]\\\\n.\\\\nWe are interested in the competitive ratio of our online algorithm A compared to the offline\\\\nalgorithm OPT . In this setting there are two different offline algorithms to compare against:\\\\noptimal offline and sequential offline. They both know S,B, but the first can always achieve\\\\nthe maximum welfare, whereas the second operates under the same constrains as we, namely\\\\nhe can only perform trades permitted by σ, which is unknown. We say that algorithm A is\\\\nρ-competitive for welfare (or gain from trade) if for any B,S we have:\\\\nWA(B,S) ≥ ρ · WOPT(S,B)− α, (1)\\\\nfor some fixed α ≥ 0.\\\\nOften we will refer to the matching between a set of buyers and a set of sellers. Let\\\\nM(S,B) = {{S1} ∪ {B1}}, where S1 ⊆ S,B1 ⊆ B is the set of sellers and buyers with whom\\\\nwe trade (or are matched, in the sense that the items move from sellers to buyers) in a wel-\\\\nfare maximising allocation and m(S,B) the optimal gain from trade. Note that this does\\\\nnot contain pairs: only the set of each side of the matching. Similarly, let M(S,B, q, p) be\\\\nthe matching generated by only trading with sellers valued below q and buyers above p. In\\\\na slight abuse of notation, we will use |M(S,B)| = |S1| for the size of the matching and\\\\nM(S,B) ?M(S′, B′) = {{S1 ? S′1} ∪ {B1 ? B′1}}, where ? is any set operation. For convenience,\\\\nwe refer to M(σ) =M({s | s ∈ σ} , {b | b ∈ σ} where σ is a sequence of agents.\\\\n3 Welfare\\\\nIn order to approximate the welfare, the online algorithm uses a sampling phase to find the\\\\nmedian price, in an attempt to transfer items from agents below the median to more valuable\\\\nones above it. The two main challenges, in terms of its performance, are estimating the median\\\\nwith a small sample and not missing too many trades due to the online nature of the input.\\\\nBefore we delve into the actual algorithm, it is useful to state two probability concentration\\\\nresults, similar to the familiar Azuma-Hoeffding inequality, but for the setting where sampling\\\\nhappens without replacement as is our case.\\\\nLemma 1. Let X = {x1, . . . , xN} where xi ∈ {0, 1}, x1 = x2 = . . . = xm = 1 and xm+1 =\\\\n. . . = xN = 0 for some integer m ≥ 0. Consider sampling n values of X uniformly at random\\\\nwithout replacement and let Xi be the value of the i− th draw. For Y =∑ni=1Xi, we have that\\\\nfor any \\\\x0f > 0:\\\\nPr[Y ≥ (1 + \\\\x0f)E [Y ]] ≤ e−2\\\\x0f2max{m,n}mnN2 (2)\\\\nand\\\\nPr[Y ≤ (1− \\\\x0f)E [Y ]] ≤ e−2\\\\x0f2max{m,n}mnN2 . (3)\\\\n5\\\\nProof. Let Yi = E [Y |X1, . . . , Xi] be the Doob martingale of Y , exposing the choices of the first\\\\ni draws. Clearly we have that |Yi+1 − Yi| ≤ 1, since the knowledge of one draw cannot change\\\\nthe expectation by more than 1. Applying Azuma’s inequality, we obtain:\\\\nPr[Yn − Y0 ≥ t] ≤ e\\\\n−t2\\\\nn . (4)\\\\nLet Zj for 1 ≤ j ≤ m indicate if xj was chosen. Since only these xj contribute to Y , we\\\\nhave that Y =∑mi=1 Zi. Repeating the previous martingale construction, we get:\\\\nPr[Ym − Y0 ≥ t] ≤ e\\\\n−t2\\\\nm . (5)\\\\nBut, we know that Y0 = E [Y ] = E [\\\\n∑m\\\\ni=1 Zi] = m nN . Setting t = \\\\x0fm\\\\nn\\\\nN in both (4) and (5)\\\\nand using Yn = Ym = Y we obtain:\\\\nPr[Y ≥ (1 + \\\\x0f)E [Y ]] ≤ e−2\\\\x0f2max{m,n}mnN2 . (6)\\\\nConcentration in the opposite direction is found by repeating the same analysis, using the\\\\ncomplementary form of Azuma’s inequality.\\\\nNote that this result is not superfluous: by immediately applying Hoeffding’s inequality for\\\\nsampling with replacement, we would obtain:\\\\nPr[Y ≥ (1 + \\\\x0f)E [Y ]] ≤ e−2\\\\x0f2m\\\\n2n\\\\nN2 ,\\\\nwhich is only tight if m is large compared to N . The concentration should intuitively work if n\\\\nis a large fraction of N as well: imagine n = N .\\\\nSimilarly, we often encounter a situation where we are interested in the number of trades\\\\nbetween n sellers and n buyers, arriving in a uniformly random permutation. Assuming we buy\\\\nfrom all sellers, occasionally we would encounter a buyer without having any items at hand.\\\\nThis \\\\nresults shows that even though this is the case, few trades are lost.\\\\nLemma 2. The number of trades M(σ), where σ is a uniformly random sequence containing\\\\nn buyers and n sellers, is:\\\\nE [M(σ)] ≥ n− 1\\\\nn\\\\n(\\\\nn−√2n logn) , (7)\\\\nassuming all sellers are valued below all buyers.\\\\nProof of Lemma 2. Since we buy from all sellers and attempt to sell to all buyers, let Xt be 1\\\\nif at step t a seller is encountered and −1 if it is a buyer. We define the following martingale,\\\\nwith Y0 = 0:\\\\nYt+1 =\\\\n\\\\uf8f1\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f2\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f3\\\\nYt +Xt if Yt > 0\\\\nYt −Xt if Yt < 0{\\\\n1 with probability 12\\\\n−1 with probability 12\\\\nif Xt = 1 and Yt = 0\\\\nYt otherwise\\\\n.\\\\nBasically, Yt keeps track of the unsold items: sellers pull away from 0 and buyers towards 0.\\\\nWe need to define Yt since Xt is not a martingale: E [Xt|Xt−1 = 0] > 0. Inductively, it is easy\\\\n6\\\\nto show that |Yt| = Xt. Therefore, the number, of unsold items at time t is at most |Yt|. By a\\\\nsimple case analysis we have that |Yt+1 − Yt| < 1. Thus, by Azuma’s inequality we have:\\\\nPr[|Y2n − Y0| ≥\\\\n√\\\\n2n logn] ≤ e−2 4n logn2·2n = 1\\\\nn\\\\n. (8)\\\\nSince n items are bought, we have:\\\\nE [M(σ)] ≥ n− 1\\\\nn\\\\n(\\\\nn−√2n logn) (9)\\\\nAll the machinery is now in place analyse sequential algorithms in this setting. We first\\\\nshow a key property of the offline algorithm.\\\\nProposition 1. The optimal offline algorithm sets a price p, equal to the median of all the\\\\nagents’ valuations and trades items from sellers valued below p to buyers valued above p.\\\\nProof. Since there are only n items available, if we could freely redistribute the items we would\\\\nchoose the top n agents with highest valuations. Let p be the value of the n-th most valuable\\\\nagent. If there are k buyers valued above p we have n − k buyers and k sellers valued below\\\\nit. Thus, buying from all sellers below p and selling to all buyers above it is an optimal\\\\nalgorithm.\\\\nHowever, the optimal sequential offline algorithm would not just trade at this price. For\\\\ninstance, if there is 1 buyer and n−1 sellers above p and 1 seller and n−1 buyers below, trading at\\\\nthis price would give a 1/2 probability of transferring the item, since only one transfer increases\\\\nthe welfare and the agents have to appear in the right order. Therefore, if that buyer has a\\\\nmuch larger valuation than anyone else, this algorithm would only be 1/2-competitive. However,\\\\nwe can modify this approach with a bias towards buying more items than needed, in order to\\\\nmaximise the probability of finding high valued buyers.\\\\nLemma 3. The optimal sequential online algorithm is\\\\n(\\\\n1−O\\\\n(\\\\nlogn\\\\nn1/3\\\\n))\\\\n-competitive against the\\\\noptimal offline for welfare.\\\\nProof. The optimal online algorithm adjusts the price p according to the following two cases,\\\\nwhere M = |M(S,B)|.\\\\n1. M ≥ n2/3. In this case the same price p is used. At the end, the online algorithm will\\\\nstill keep the highest valued n−M sellers and by Lemma 2 will match\\\\nM − 1\\\\nM\\\\n(\\\\nM −√2M logM)\\\\nbuyers in expectation. The offline optimum will of course keep the highest n−M sellers\\\\nand M buyers, leading to a competitive ratio of at most:\\\\nM − 1\\\\nM\\\\n(\\\\n1−\\\\n√\\\\n2 logM√\\\\nM\\\\n)\\\\n= 1−O\\\\n( logn\\\\nn1/3\\\\n)\\\\n.\\\\n2. M < n2/3.\\\\nIn this case, suppose two prices are used: pS to buy from the lowest n2/3 sellers and pB\\\\nto sell to the highest n2/3 buyers. For the buyers, the online does at least as well as\\\\n7\\\\nthe previous case. In particular, it it obtains a uniformly random sample of size at least\\\\nn2/3−\\\\n√\\\\n2n2/3 logn2/3 by Lemma 2, amongst the top n2/3 buyers with probability at least\\\\n(n2/3 − 1)/n2/3. Since the M buyers matched by the optimal offline are contained within\\\\nthe highest n2/3 buyers, the ratio just from buyers remains the same as before.\\\\nFrom the sellers side, the online keeps the highest n− n2/3 sellers, while the offline keeps\\\\nat most n, for a ratio at most 1− 1/n1/3.\\\\nCombining both cases, the ratio is asymptotically at most:\\\\n1−O\\\\n( logn\\\\nn1/3\\\\n)\\\\n. (10)\\\\nNote that the choice of n2/3 to separate the two cases is optimal.\\\\nThe next step is to design an online algorithm without knowing p or |M(S,B)| beforehand.\\\\nThe algorithm is as follows:\\\\n1. Record the first 8n2/3 logn agents and calculate their median p′. Buy from all sellers\\\\nduring this sampling phase.\\\\n2. After the sampling starts the trading phase:\\\\n(a) Buy from seller s if s ≤ p′.\\\\n(b) Sell to buyer b is an item is available and b ≥ p′.\\\\nFor the analysis of this algorithm, we first need a concentration result on the sample median p′.\\\\nLemma 4. Let X = {1, . . . , 2n} and select 8n2/3 logn elements from X without replacement.\\\\nThen, their sample median M satisfies:\\\\nPr[|M − n| ≥ n2/3] ≤ O\\\\n( 1\\\\nn\\\\n)\\\\n. (11)\\\\nProof. We have that:\\\\nPr[M ≥ n+ n2/3] = Pr[more than 4n2/3 logn elements picked no less than n+ n2/3]\\\\nSince we are sampling without replacement, this is equivalent to selecting 8n2/3 logn elements\\\\nuniformly at random from X ′ containing n + n2/3 0’s and n − n2/3 1’s and having their sum\\\\nbe greater than 4n2/3 logn. Using X ′, \\\\x0f = n2/3/(n − n2/3) and taking 8n2/3 logn samples in\\\\nLemma 1, we have:\\\\nPr[M ≥ n+ n2/3] = Pr[Y ≥ (1 + \\\\x0f)E [Y ]] = Pr[Y ≥ n2/3]\\\\n≤ exp\\\\n\\\\uf8eb\\\\uf8ed−2( n2/3\\\\n(n− n2/3)\\\\n)2\\\\n(n− n2/3)8n\\\\n2/3 logn(n− n2/3)\\\\n4n2\\\\n\\\\uf8f6\\\\uf8f8\\\\n≤ O\\\\n( 1\\\\nn\\\\n)\\\\n.\\\\nBy symmetry, the same holds for Pr[M ≤ n−n2/3]: just reverse the ordering of the agents.\\\\nThis shows that our sample median p′ might have at most n2/3 agents more on one side\\\\ncompared to the true median p. However, this loss is negligible asymptotically, as these agents\\\\nare a uniformly random subset of the S ∪B. We now show that buying from sellers during the\\\\nsampling phase, before considering any buyers, can only increase the number of trades in the\\\\nnext phase.\\\\n8\\\\nLemma 5. Let σ be a sequence containing n buyers and n sellers. Move an arbitrary seller the\\\\nbeginning of the sequence to obtain sσ′. Then we have:\\\\n|M(sσ′)| ≥ |M(σ)|.\\\\nProof. Let b′ be the first buyer not to receive an item in σ. Clearly, if b′ doesn’t exist then the\\\\nnumber of items sold in both cases is n. Assume we sell the item bough from s only if it is the\\\\nlast item left. Then, it is sold to b′: otherwise b′ would not be the first buyer not to be sold an\\\\nitem is σ. There are two cases:\\\\n1. If s appears in σ before b′: both sequences continue identically as we have no items in\\\\nstock after b′.\\\\n2. If s appears after b′ in σ: there is one fewer seller in sσ′ after b′, since s was moved to the\\\\nfront. However, this can result in at one lost sale.\\\\nActually, we have shown that moving sellers to the beginning can only increase trades, which\\\\nis slightly more powerful. We are now ready to state one of the main \\\\nresults of this section.\\\\nTheorem 1. This algorithm is\\\\n(\\\\n1− O˜\\\\n(\\\\n1\\\\nn1/3\\\\n))\\\\n-competitive for welfare.\\\\nProof. As before, let M = |M(S,B)| be the size of the optimal offline matching. The following\\\\nanalysis assumes that the event of Lemma 4 did not occur and p and p′ split the agents in two\\\\nsets, differing by at most n2/3. Given this, we analyse the algorithm in three steps. First show\\\\nthat we never buy too many items from highly valued sellers, therefore we keep most of the\\\\nsellers’ contribution to the final welfare. Then we show that we always match a high proportion\\\\nof the valuable buyers by considering two cases: if there are few such buyers then they are\\\\nmatched to the sellers we obtained during the sampling phase, otherwise we have enough sellers\\\\nbelow p′ to match them to.\\\\nWe introduce some notation useful to the analysis: let W be the set containing the top\\\\nn−n2/3 highest valued agents. Then let SW , BW be the number of sellers and buyers respectively\\\\nin W and S′W , B′W be how many of them appeared after the sampling phase. valued agents. To\\\\nshow the competitiveness of our algorithm, it suffices to find the fraction of W that is achieved\\\\nat the end of the sequence: being (1− O˜(1/n1/3))-competitive against the top n− n2/3 agents\\\\nimplies a ratio of (\\\\n1− O˜(1/n1/3)\\\\n)\\\\n· n− n\\\\n2/3\\\\nn\\\\n= 1− O˜(1/n2/3)\\\\nagainst all n agents above the median and therefore the optimal offline.\\\\nWe first show that we never lose too much welfare by buying from sellers, both in\\\\nthe sampling and trading phase. Given p′, the only occasion on which a seller in W is\\\\nbought is if he is amongst the first 8n2/3 logn sellers. This event is clearly independent from\\\\nthe condition on p′, meaning in expectation we keep\\\\nE\\\\n[\\\\nS′W\\\\n]\\\\n= SW\\\\n(\\\\n1− 8n\\\\n2/3 logn\\\\nn\\\\n)\\\\n= SW\\\\n(\\\\n1− 8 logn\\\\nn1/3\\\\n)\\\\n(12)\\\\nhighly valued sellers. Therefore, enough of the sellers’ original value is kept. The rest of the\\\\nanalysis will only focus only the proportion of buyers in W who get an item. For the number\\\\nof items IS bought during the sampling phase, the following holds by Lemma 1:\\\\nPr\\\\n[\\\\nIS ≤ (1− 12)4n\\\\n2/3 logn\\\\n]\\\\n≤ e−2 148n2/3 logn n\\\\n2\\\\n4n2 ≤ e−n2/3 , (13)\\\\n9\\\\nas there are n out of 2n agents are sellers and we sample 8n2/3 logn of them. Therefore, we\\\\nenter the trading phase with an excess of at least 2n2/3 logn items with high probability.\\\\nTo analyse the number of buyers in W matched, we consider two cases.\\\\nBW ≤ n2/3 logn: In this case there are few valuable buyers and all we need to show is\\\\nthat the excess of items bought during sampling is enough to trade with most of them. We\\\\nfirst need to find E [B′W ], which is slightly more complicated, since we have conditioned on p′\\\\napproximating the median. Given p′, at least 4n2/3 logn agents were above the median value\\\\nduring the sampling phase. Note that all of the agents in W are above the median. Therefore,\\\\nany of the agents in the upper 4n2/3 logn half of the sampling phase could be replaced by a\\\\nbuyer in W . At worst, 4n2/3 logn agents from W are in the sampling phase, which means that\\\\nin a random permutation, we have:\\\\nn2/3 logn ≥ E [B′W ] ≥ BW\\\\n(\\\\n1− 4n\\\\n2/3 logn\\\\nn− n2/3\\\\n)\\\\n.\\\\nWe might also consider up to n2/3 extra buyers, if p′ underestimated p. However, given that\\\\nIS ≥ 2n2/3 logn with high probability, every buyer in B′W will be matched with an item, giving\\\\nthe claimed competitive ratio for this case.\\\\nBW > n\\\\n2/3 logn: Let k ≥ BW be the number of trades the optimal offline algorithm would\\\\nperform. Since the median might be underestimated, the number of sellers we consider is at\\\\nleast k − n2/3 and buyers at most k + n2/3. We show that, with the help of the extra items we\\\\nbought during sampling, we have more items than buyers in total, with high probability. Let\\\\nSp′ , Bp′ the number of sellers and buyers below and above p′ after the sampling phase. By by\\\\nLemma 1 we expect to find\\\\nPr[Sp′ ≤ (1−\\\\n√\\\\nlog k\\\\nk\\\\n)(k − n2/3)2n− 8n\\\\n2/3 logn\\\\n2n ] (14)\\\\n≤ exp\\\\n(\\\\n−2log k\\\\nk\\\\n(k − n2/3)(2n− 8n\\\\n2/3 logn)2\\\\n4n2\\\\n)\\\\n≤ O\\\\n(1\\\\nk\\\\n)\\\\n, (15)\\\\nby sampling 2n− 8n2/3 logn out of 2n with k − n2/3 important elements. Similarly we have\\\\nPr[Bp′ ≥ (1 +\\\\n√\\\\nlog k\\\\nk\\\\n)(k + n2/3)2n− 8n\\\\n2/3 logn\\\\n2n ] ≤ O\\\\n(1\\\\nk\\\\n)\\\\n. (16)\\\\nIt is important to note that these quantities are almost equal, other than a n2/3 factor which is\\\\ninsignificant compared to k. Then, with high probability (well O(1/k), relatively high):\\\\nBp′ − Sp′ ≤ (1 +\\\\n√\\\\nlog k\\\\nk\\\\n)(k + n2/3)− (1−\\\\n√\\\\nlog k\\\\nk\\\\n)(k − n2/3) (17)\\\\n≤ 2(√k log k + n2/3) (18)\\\\n≤ 3(√n logn), (19)\\\\ngiven than k ≤ n. Since we bought at least 2n2/3 logn items during the sampling phase, the\\\\ntotal number of items bought is higher than the total number of buyers considered for n large\\\\nenough. Also, by Lemma 5 having these items ready before encounter buyers is beneficial.\\\\n10\\\\nTherefore, we get a lower bound on the number of buyers in Bp′ that actually acquire an\\\\nitem using Lemma 2. The number of items sold in expectation is at least:\\\\nM≥ Bp′ − 1\\\\nBp′\\\\n(\\\\nBp′ −\\\\n√\\\\n2Bp′ logBp′\\\\n)\\\\n. (20)\\\\nHowever, we are interested only in the fraction of buyers in BW who acquired an item. The\\\\nalgorithm does not differentiate between any buyer above p′, the sequence is uniformly random\\\\nand all buyers in BW are contained within the top k+n2/3 buyers. By lower bounding Bp′ with\\\\nLemma 1:\\\\nPr[Bp′ ≤ (1−\\\\n√\\\\nlog k\\\\nk\\\\n)(k + n2/3)2n− 2n\\\\n2/3 logn\\\\n2n ] ≤ O\\\\n(1\\\\nk\\\\n)\\\\n, (21)\\\\nand using (20), the fraction of buyers in BW matched is at least:\\\\nM\\\\nk + n2/3\\\\n≥ 1− O˜\\\\n( 1\\\\nn2/3\\\\n)\\\\n, (22)\\\\nwith probability 1−O(1/k), which is asymptotically high as k ≥ n2/3 logn.\\\\n4 Gain from Trade\\\\nCompared to the welfare, the gain from trade is a more challenging objective. The main reason\\\\nis that even for large n, the actual trades that maximise the GFT can be very few and quite\\\\nwell hidden. Moreover, buying from a single seller and being unable to sell could completely\\\\nshatter the GFT, while it could have very little effect on the welfare.\\\\nFirst of all, the setting has to be slightly changed. We give the online algorithm one extra,\\\\nfree item at the beginning to ensure that at least one buyer can acquire an item, even when the\\\\ninitial sampling has been inconclusive. For fairness, the offline algorithm is also provided with\\\\nthis starting item. We show that this modification is absolutely necessary to study this setting\\\\nunder competitive analysis.\\\\nTheorem 2. Starting with no items, there exist S,B such that the competitive ratio for the\\\\nGFT is arbitrarily high.\\\\nProof. Consider two different valuations. The first has s1 = c > 0 and b1 = c+ \\\\x0f. In the second\\\\nhas sˆ1 = c, bˆ1 = c − \\\\x0f, sˆ2 < bˆ1 − \\\\x0f′. We tweak the value of the buyer so that the trade from\\\\ninstance one no longer increases welfare, but add one extra seller to keep the optimal GFT\\\\npositive.\\\\nLet p = Pr[s1 ∈ TS | σ−1(b1) > σ−1(s1)] be the probability of the online algorithm buying\\\\nfrom s1, conditioned on b1 arriving later. This must be p > 0, otherwise his expected gain from\\\\ntrade will be 0, compared to the \\\\x0f/2 generated by the offline.\\\\nHowever, in the second instance the algorithm should buy from sˆ2 instead of sˆ1. But, if sˆ1\\\\nappears first, the first algorithm should buy from him too, as the information received so far is\\\\nthe same:\\\\npˆ = Pr[sˆ1 ∈ TS | σ−1(bˆ1) > σ−1(sˆ1) ∧ σ−1(sˆ2) > σ−1(sˆ1)]\\\\n≥ pPr[σ−1(sˆ2) > σ−1(sˆ1)]\\\\n= p2 > 0.\\\\n11\\\\nSo the online algorithm has a positive chance of buying the item from the wrong seller. Assuming\\\\nin all other cases maximum gain from trade is extracted, we have:\\\\nGFTA(S,B) ≤ pˆ(−\\\\x0f) + (1− pˆ)(\\\\x0f′ − \\\\x0f). (23)\\\\nSince pˆ is independent of \\\\x0f, \\\\x0f′, we can set\\\\n\\\\x0f′ = \\\\x0f1− pˆ\\\\nwhich leads to GFTA(S,B) = 0 whereas the offline has \\\\x0f/2.\\\\nIn any case, no online algorithm can perform well in both instances.\\\\nTo avoid the previous pitfall, we assume the intermediary starts with one item. Roughly, the\\\\nalgorithms starts by estimating the total volume of trades in an optimal matching by observing\\\\nthe first segment of the sequence. Using this information, two prices pˆ, qˆ are computed, to be\\\\noffered to agents in the second part. This being an ordinal mechanism, the goal is to maximise\\\\nthe number of trades and leave no item unsold. During the trading phase we are also much\\\\nmore conservative: at most one item is kept in stock and we stop buying items well before the\\\\nend of the sequence, to make sure that there are enough buyers left to sell everything. The\\\\nonline algorithm A(c, \\\\x0f,N) contains parameters whose values will be specified later.\\\\nInput: A sequence σ of length 2n, appearing online.\\\\nOutput: A matching between buyers and sellers.\\\\nWith probability 12 ignore sellers and sell the item as in the normal secretary, otherwise\\\\ncontinue ;\\\\nSplit the sequence into two segments such that σ = σ1σ2, with |σ1| = c · 2n;\\\\nLet S1, B1 denote the sellers and buyers of σ1;\\\\nCalculate the welfare maximising matching M(S1, B1);\\\\nif |M(S1, B1)| ≤ N then\\\\nSell the item to the highest buyer as in the normal secretary problem and stop;\\\\nend\\\\nSet pˆ, qˆ which only keep (1− \\\\x0f) · c · |M(S1, B1)| many matched pairs;\\\\ni← c · 2n;\\\\nk ← ∅;\\\\nM ← ∅;\\\\n/* For the first half of σ2, buy and sell items, keeping at most one in\\\\nstock */\\\\nwhile i ≤ c · 2n+ (1− c) · 2n/2 do\\\\nif σ(i) is a seller, k = ∅ and σ(i) ≤ qˆ then\\\\nk ← σ(i);\\\\nend\\\\nif σ(i) is a buyer, k 6= ∅ and σ(i) ≥ pˆ then\\\\nSell to σ(i);\\\\nk ← ∅;\\\\nend\\\\ni← i+ 1;\\\\nend\\\\nFor the second half of σ2, just try to sell the last remaining item, if any;\\\\nThe idea is to use the first part of the sequence to estimate the matching M(S,B). If a\\\\nlarge (in terms of pairs) GFT maximising matching is observed, it is likely that a proportionate\\\\n12\\\\nfraction of it will be contained in the second half. In that case, sellers and buyers are matched\\\\nin non overlapping pairs to avoid buying too many items. However, if the observed matching is\\\\ntoo small, then the algorithm defaults to selling only the starting item, as it is very likely that\\\\nσ2 will not contain enough buyers for anything more.\\\\nBefore moving on to the analysis of the algorithm, we need a simple lemma on the structure\\\\nof GFT maximising matchings, to explain the prices set.\\\\nLemma 6. For any S,B and S1 ⊆ S,B1 ⊆ B:\\\\n1. m(S,B) can be obtained by setting two threshold prices p, q and trading with buyers above\\\\nand sellers below them.\\\\n2. Choosing pˆ > p and qˆ < q such that |M(S,B, qˆ, pˆ)| ≥ α|M(S,B)| for α < 1 yields\\\\nm(S,B, qˆ, pˆ) ≥ αm(S,B).\\\\n3. |M(S,B)| ≥ |M(S1, B1)| and m(S,B) ≥ m(S1, B1).\\\\nProof. For Property 1, assume s < b < sˆ < bˆ, such that s, b, sˆ, bˆ ∈M(S,B). But, instead of two\\\\nmatches we can just match bˆ to s instead: bˆ− s > b− s+ bˆ− sˆ, thus any such pair of matched\\\\nagents cannot be part of M(S,B). Setting q = max {s ∈M(S,B)} and p = min {b ∈M(S,B)}\\\\nwe have q < p and the result follows. This is essentially the same observation as using the\\\\nmedian price to trade, but using two different prices for robustness, as we will see later.\\\\nProperty 2 follows because M(S,B, qˆ, pˆ) contains the α highest value pairs for M(S,B).\\\\nProperty 3 is straightforward.\\\\nTheorem 3. A(c = 0.3, \\\\x0f = 0.2758, N = 114) is O(1)-competitive for the gain from trade.\\\\nProof. Let z = |M(S,B)|. We bound the gain from trade for the case where σ1, σ2 contain\\\\ntheir analogous proportion of M(S,B) and show that the losses are insignificant otherwise. In\\\\nparticular, let\\\\nf(c, \\\\x0f, z)\\\\n= Pr\\\\n[\\\\nM(S,B) ∩M(S1, B1)\\\\nM(S,B) ≥ c(1− \\\\x0f) ∧\\\\nM(S,B) ∩M(S2, B2)\\\\nM(S,B) ≥ (1− c)(1− \\\\x0f)\\\\n]\\\\nbe the well mixed probability, where an \\\\x0f-approximate chunk of the matching appears in both\\\\nparts. The two events are not independent. To bound f(c, \\\\x0f, z), it suffices to study the distri-\\\\nbution of SM = {s ∈M(S,B)} and BM = {b ∈M(S,B)}, the sets of agents comprising the\\\\noptimal matching. By Lemma 6, we know that any seller in SM can be matched to any buyer\\\\nin BM . Since we only care about the size of the matching in σ1 and σ2, not its actual value, we\\\\ncan rewrite f(c, \\\\x0f, z) as:\\\\nf(c, \\\\x0f, z) = Pr\\\\n[\\\\n|SM ∩ S1|\\\\n|SM | ≥ c(1− \\\\x0f) ∧\\\\n|BM ∩B1|\\\\n|BM | ≥ c(1− \\\\x0f)∧ (24)\\\\n|SM ∩ S2|\\\\n|SM | ≥ (1− c)(1− \\\\x0f) ∧\\\\n|BM ∩B2|\\\\n|BM | ≥ (1− c)(1− \\\\x0f)\\\\n]\\\\n, (25)\\\\nwhich is easier to handle.\\\\nIt is useful to think the input as being created in two steps: first the volume of agents in\\\\nS1, B1, S2, B2 is chosen and afterwards their exact values are randomly assigned. As such, a\\\\nlower bound on the fraction of the size of the online to the offline matching provides the same\\\\nbound on the gain from trade. We begin by bounding f(c, \\\\x0f, z).\\\\n13\\\\nLemma 7. The probability the matching is well-mixed is\\\\nf(c, \\\\x0f, z) ≥ 1− 2(e−2\\\\x0f2zc2 + e−2\\\\x0f2z(1−c)2)\\\\nProof. Continuing from (24) we have:\\\\nf(c, \\\\x0f, z) = Pr\\\\n[\\\\n|SM ∩ S1|\\\\n|SM | ≥ c(1− \\\\x0f) ∧\\\\n|BM ∩B1|\\\\n|BM | ≥ c(1− \\\\x0f)∧\\\\n|SM ∩ S2|\\\\n|SM | ≥ (1− c)(1− \\\\x0f) ∧\\\\n|BM ∩B2|\\\\n|BM | ≥ (1− c)(1− \\\\x0f)\\\\n]\\\\n≥ 1− Pr\\\\n[ |SM ∩ S1|\\\\n|SM | ≤ c(1− \\\\x0f)\\\\n]\\\\n− Pr\\\\n[ |BM ∩B1|\\\\n|BM | ≤ c(1− \\\\x0f)\\\\n]\\\\n−\\\\nPr\\\\n[ |SM ∩ S2|\\\\n|SM | ≤ (1− c)(1− \\\\x0f)\\\\n]\\\\n− Pr\\\\n[ |BM ∩B2|\\\\n|BM | ≤ (1− c)(1− \\\\x0f)\\\\n]\\\\n(26)\\\\n≥ 1− 2(e−2\\\\x0f2zc2 + e−2\\\\x0f2z(1−c)2), (27)\\\\nwhere (26) follows by taking the complement and a union bound and (27) by applying Lemma 1\\\\nindividually for each event.\\\\nLet p and q be the prices achieving the matching M(S,B), by Lemma 6. We need to show\\\\nthat the prices pˆ, qˆ computed achieve a constant approximation of m(S2, B2). Since M(S,B) is\\\\nwell mixed and by using Lemma 6 we have that:\\\\n|M(S,B)| ≥ |M(S1, B1)| ≥ |M(S1, B1, q, p)| ≥ (1− \\\\x0f) · c · |M(S,B)|, (28)\\\\nwhere the second inequality holds sinceM(S1, B1) is a gain from trade maximising matching and\\\\nthe third because at least a (1− \\\\x0f) · c fraction of M(S,B) appeared in σ1. In particular, we have\\\\nthat M(S1, B1, q, p) ⊆ M(S1, B1) is the highest value part of M(S1, B1) and M(S1, B1, qˆ, pˆ) ⊆\\\\nM(S1, B1, q, p), thus qˆ ≤ q and pˆ ≥ p leading to:\\\\n|M(S1, B1, qˆ, pˆ)| ≥ (1− \\\\x0f)2c2|M(S,B)| (29)\\\\nby Eq. (28). Therefore, the prices pˆ, qˆ computed find a relatively large subset of M(S,B). We\\\\nnow need to find just how many of the trades in M(S2, B2, pˆ, qˆ) are achieved by our algorithm.\\\\nLet Sˆ2 = {s | s ∈ S2 ∧ s < qˆ} and Bˆ2 = {b | b ∈ B2 ∧ b > pˆ}. We need a high probability\\\\nguarantee on the size of Sˆ2 and Bˆ2.\\\\nLemma 8. Assuming the matching is well mixed:\\\\nPr\\\\n[\\\\n|Sˆ2| ≥ ((1− c)(1− \\\\x0f)− 12)|SM |\\\\n]\\\\n≥ 1− 2−c2(1−\\\\x0f)2|SM |.\\\\nProof. In the well-mixed case, we have that\\\\n|S2 ∩ SM | ≥ (1− c)(1− \\\\x0f)|SM | and |S1 ∩ SM | ≥ c(1− \\\\x0f)|SM | (30)\\\\nwhich leads to\\\\n|S1 ∩ SM | ≤ (1− (1− c)(1− \\\\x0f))|SM |. (31)\\\\nTo get a lower bound on the size, we have:\\\\nPr\\\\n[\\\\n|Sˆ2| ≥ |SM |2 − |S1 ∩ SM |\\\\n]\\\\n≥ Pr\\\\n[\\\\n|Sˆ2| ≥ ((1− c)(1− \\\\x0f)− 12)|SM |\\\\n]\\\\n(32)\\\\n≥ Pr [qˆ ≥ median(SM )] (33)\\\\n≥ 1− 2−c2(1−\\\\x0f)2|SM | (34)\\\\n14\\\\nwhere Eq. (32) follows from Eq. (30). Eq. (33) follows since if qˆ is greater than the median, then\\\\nat worst case all elements from S1 ∩ SM are less than qˆ, which still leaves plenty of sellers in\\\\nS2. Eq. (34) follows since draws are not actually independent, but this works in the inequality’s\\\\nfavour. From Eq. (29) we know qˆ is greater than at least a c2(1− \\\\x0f)2 fraction of sellers. Since\\\\nthe ‘bad case’ is choosing all sellers below the median, this happens with higher probability if\\\\neach draw is with rather than without replacement, leading to the result.\\\\nClearly, Lemma 8 holds for buyers as well. The proof is almost identical, keeping in mind that\\\\nbuyers are ordered the opposite way.\\\\nAt this point we have a clear indication of how many sellers and buyers the prices pˆ, qˆ cover\\\\nin the second part of the sequence. Since this is an ordinal mechanism, we want to maximise\\\\nthe number of trades provided no item is left unsold. There are no a priori guarantees on the\\\\nwelfare increase of each trade, even a single unsold item ruins our gain from trade guarantees,\\\\nin the worst case.\\\\nLemma 9. Let A = |M(S2, B2, qˆ, pˆ)| and B = |S2| + |B2| − A. Then, the probability that no\\\\nitem is left unsold is at least 1 − 2−A. Moreover, the expected number of trades in this case is\\\\nat least:\\\\nA+B/2−1\\\\n2A+B−1 · A2 − A2A\\\\n1− 2−A ≈\\\\n|M(S2, B2, qˆ, pˆ)|\\\\n4 . (35)\\\\nProof. We begin by calculating the probability of having an unsold item, which is easy: it is\\\\nat most as much as the probability of not encountering a buyer within the last (1 − c) · 2n/2\\\\nagents. Using a similar argument as Lemma 8, this probability is at most 2−A.\\\\nWe now need to calculate the expected number of trades. LetXi be a random variable indicating\\\\nthat an item was sold to the i-th agent. We have:\\\\nPr [Xi = 1] = Pr [ previous transaction was buying ∧Xi is a buyer ]\\\\n= A2A ·\\\\nA\\\\n2A+B − 1 =\\\\n1\\\\n2 ·\\\\nA\\\\n2A+B − 1 ,\\\\nsince the previous transaction being buying from a seller occurs with probability A2A as there\\\\nare A sellers in M(S2, B2, pˆ, qˆ) for 2A total agents and the sequence is shuffled. The second\\\\nfraction has 2A + B − 1 for the denominator, taking into account that one seller has already\\\\nbeen used.\\\\nBy linearity of expectation, the total number of trades X is (note that we only consider the\\\\nfirst half of σ2, where we both buy and sell):\\\\nE [X] ≥ E\\\\n\\\\uf8ee\\\\uf8f0(2A+B)/2∑\\\\ni=2\\\\nXi\\\\n\\\\uf8f9\\\\uf8fb = (A+B/2− 1)E [Xi]\\\\n= (A+B/2− 1)12 ·\\\\nA\\\\n2A+B − 1\\\\n= A+B/2− 12A+B − 1 ·\\\\nA\\\\n2\\\\n15\\\\nWe can use this to calculate the expected number of trades in the case where nothing is left\\\\nunsold:\\\\nE [X|No items unsold] = E [X]− E [X| Unsold items] Pr [Unsold Items]Pr [No unsold items]\\\\n≥ E [X]−\\\\nA\\\\n2−A\\\\n1− 2−A\\\\nEverything is now in place to provide a lower bound on the gain from trade of the matching\\\\ncalculated by the algorithm. Assuming z = |M(S,B)|, we can compose Lemma 7, Lemma 8\\\\nand Lemma 9 to show that with probability at least\\\\nJ(c, \\\\x0f, z) = f(c, \\\\x0f, z) · (1− 2−c2(1−\\\\x0f)2z) · (1− 2−((1−c)(1−\\\\x0f)− 12 )z), (36)\\\\nthe matching has size at least\\\\n((1− c)(1− \\\\x0f)− 12)z\\\\n4 . (37)\\\\nThe matching is not a uniformly random subset of M(S,B), but it is skewed to contain higher\\\\nvalue trades since pˆ > p and qˆ < q. Taking into account that we run a simple secretary algorithm\\\\nwith probability 1/2 and assuming we lose the highest valued seller s? in our matching when\\\\nthe agents are not well mixed (we can only have one unsold item) the GFT is at least:\\\\n1\\\\n2eb\\\\n1 + J(c, \\\\x0f, z)2 ·\\\\n((1− c)(1− \\\\x0f)− 12)m(S,B)\\\\n4 −\\\\n1− J(c, \\\\x0f, z)\\\\n2 s\\\\n? (38)\\\\nwhereas the offline GFT is at most\\\\nm(S,B) + b1. (39)\\\\nTo upper bound the competitive ratio ρ we analyse three different cases:\\\\n1. If z < N :\\\\nIn this case the algorithm would never detect a sufficiently sized matching and would\\\\nalways run a simple secretary algorithm. Note this is possible, as c = 0.3 ≤ 1/e required\\\\nfor the secretary.\\\\nρ ≥\\\\n1\\\\neb\\\\n1\\\\nm(S,B) + b1 ≥\\\\n1\\\\ne\\\\n· b\\\\n1\\\\n(z + 1)b1 =\\\\n1\\\\ne(z + 1) (40)\\\\n2. If N ≤ z < N 1c(1−\\\\x0f) . In the well-mixed case, the online algorithm will not detect a\\\\nmatching and fall back to secretary. Therefore, the competitive ratio is:\\\\nρ ≥\\\\n1\\\\n2eb\\\\n1 + 12(f(c, \\\\x0f, z)\\\\n1\\\\neb\\\\n1 − (1− f(c, \\\\x0f, z))s?\\\\nm(S,B) + b1 ≥\\\\n1− e+ (1 + e)f(c, \\\\x0f, z)\\\\n2e(z + 1) . (41)\\\\ngiven that c < 1/e and the sampling phase for the secretary continues.\\\\n3. z ≥ N 1c(1−\\\\x0f) . Now in the well mixed case a large enough matching is found. We have:\\\\n1\\\\n2eb\\\\n1 − 1−J(c,\\\\x0f,z)2 s?\\\\nb1\\\\n≥ 12e − (1− J(c, \\\\x0f, z)), (42)\\\\nand\\\\nJ(c,\\\\x0f,z)\\\\n2 ·\\\\n((1−c)(1−\\\\x0f)− 12 )m(S,B)\\\\n4\\\\nm(S,B) ≥\\\\nJ(c, \\\\x0f, z) · ((1− c)(1− \\\\x0f)− 12)m(S,B)\\\\n8 . (43)\\\\n16\\\\nTherefore, the competitive ratio is:\\\\nρ ≥\\\\n1\\\\n2eb\\\\n1 + J(c,\\\\x0f,z)2 ·\\\\n((1−c)(1−\\\\x0f)− 12 )m(S,B)\\\\n4 − 1−J(c,\\\\x0f,z)2 s?\\\\nm(S,B) + b1\\\\n≥ min\\\\n{\\\\n1\\\\n2e − (1− J(c, \\\\x0f, z)),\\\\nJ(c, \\\\x0f, z) · ((1− c)(1− \\\\x0f)− 12)m(S,B)\\\\n8\\\\n}\\\\n.\\\\nTherefore, c, \\\\x0f and N are selected to maximise the minimum amongst all cases of z, which is\\\\npicked by the adversary. Computationally, we find that setting c = 0.3, \\\\x0f = 0.2758 and N = 114\\\\nyields ρ ≥ 1/1434.\\\\nIf we are given that |M(S,B)| will be large, then this algorithm can be adapted to have\\\\ngreatly improved competitive ratio. In particular, setting c = \\\\x0f = 0.01 achieves ρ ≥ 1/17 as\\\\n|M(S,B)| → ∞.', 'The Price of Information in Combinatorial Optimization': 'Title: The Price of Information in Combinatorial Optimization\\\\nAbstract. Suppose there are n Markov chains and we need to pay a\\\\nper-step price to advance them. The “destination” states of the Markov\\\\nchains contain rewards; however, we can only get rewards for a subset of\\\\nthem that satisfy a combinatorial constraint, e.g., at most k of them, or\\\\nthey are acyclic in an underlying graph. What strategy should we choose\\\\nto advance the Markov chains if our goal is to maximize the total reward\\\\nminus the total price that we pay?\\\\nIn this paper we introduce a Markovian price of information model to\\\\ncapture settings such as the above, where the input parameters of a\\\\ncombinatorial optimization problem are given via Markov chains. We\\\\ndesign optimal/approximation algorithms that jointly optimize the value\\\\nof the combinatorial problem and the total paid price. We also study\\\\nrobustness of our algorithms to the distribution parameters and how to\\\\nhandle the commitment constraint.\\\\nOur work brings together two classical lines of investigation: getting op-\\\\ntimal strategies for Markovian multi-armed bandits, and getting exact\\\\nand approximation algorithms for discrete optimization problems using\\\\ncombinatorial as well as linear-programming relaxation ideas.\\\\nKeywords: Multi-armed bandits · Gittins index · Probing algorithms.\\\\n1 \\\\nIntroduction\\\\nSuppose we are running an oil company and are deciding where to set up new\\\\ndrilling operations. There are several candidate sites, but the value of drilling\\\\neach site is a random variable. We must therefore inspect sites before drilling.\\\\nEach inspection gives more information about a site’s value, but the inspection\\\\nprocess is costly. Based on laws, geography, or availability of equipment, there\\\\nare constraints on which sets of drilling sites are feasible. We ask:\\\\nWhat adaptive inspection strategy should we adopt to find a feasible set\\\\nof sites to drill which maximizes, in expectation, the value of the chosen\\\\n(drilled) sites minus the total inspection cost of all sites?\\\\nLet us consider the optimization challenges in this problem:\\\\n(i) Even if we could fully inspect each site for free, choosing the best feasible\\\\nset of sites is a combinatorial optimization problem.\\\\n2 Gupta et al.\\\\n(ii) Each site may have multiple stages of inspection. The costs and possible\\\\noutcomes of later stages may depend on the outcomes of earlier stages. We\\\\nuse a Markov chain for each site to model how our knowledge about the\\\\nvalue of the site stochastically evolves with each inspection.\\\\n(iii) Since a site’s Markov chain model may not exactly match reality, we want\\\\na robust strategy that performs well even under small changes in the model\\\\nparameters.\\\\n(iv) If there is competition among several companies, it may not be possible to\\\\ndo a few stages of inspection at a given site, abandon that site’s inspection\\\\nto inspect other sites, and then later return to further inspect the first site.\\\\nIn this case the problem has additional “take it or leave it” or commitment\\\\nconstraints, which prevent interleaving inspection of multiple sites.\\\\nWhile each of the above aspects has been individually studied in the past,\\\\nno prior work addresses all of them. In particular, aspects (i) and (ii) have not\\\\nbeen simultaneously studied before. In this work we advance the state of the art\\\\nby solving the (i)-(ii)-(iii) and the (i)-(ii)-(iv) problems.\\\\nTo study aspects (i) and (ii) together, in §2 we propose the Markovian Price\\\\nof Information (Markovian PoI) model. The Markovian PoI model unifies\\\\nprior models which address (i) or (ii) alone. These prior models include those of\\\\nKleinberg et al. [33] and Singla [37], who study the combinatorial optimization\\\\naspect (i) in the so-called price of information model, in which each site has just\\\\na single stage of inspection; and those of Dimitriu et al. [17] and Kleinberg et\\\\nal. [33, Appendix G], who consider the multiple stage inspection aspect (ii) for\\\\nthe problem of selecting just a single site.\\\\nOur main \\\\nresults show how to solve combinatorial optimization problems,\\\\nincluding both maximization and minimization problems, in the Markovian\\\\nPoI model. We give two \\\\nmethods of transforming classic algorithms, originally\\\\ndesigned for the Free-Info (inspection is free) setting, into adaptive algorithms\\\\nfor the Markovian PoI setting. These adaptive algorithms respond dynami-\\\\ncally to the random outcomes of inspection.\\\\n– In §3.3 we transform “greedy” α-approximation algorithms in the Free-\\\\nInfo setting into α-approximation adaptive algorithms in the Markovian\\\\nPoI setting (Theorem 3.1). For example, this yields optimal algorithms for\\\\nmatroid optimization (Corollary 3.1).\\\\n– In §4 we show how to slightly modify our α-approximations for the Marko-\\\\nvian PoI setting in Theorem 3.1 to make them robust to small changes in\\\\nthe model parameters (Theorem 4.1).\\\\n– In §5 we use online contention resolution schemes (OCRSs) [19] to transform\\\\nLP based Free-Info maximization algorithms into adaptive Markovian\\\\nPoI algorithms while respecting the commitment constraints. Specifically,\\\\na 1/α-selectable OCRS yields α-approximation with commitment (Theo-\\\\nrem 5.1).\\\\nThe general idea behind our first result (Theorem 3.1) is the following. A\\\\nFrugal combinatorial algorithm (Definition 3.6) is, roughly speaking, “greedy”:\\\\nThe Markovian Price of Information 3\\\\nit repeatedly selects the feasible item of greatest marginal value. We show how\\\\nto adapt any Frugal algorithm to the Markovian PoI setting:\\\\n– Instead of using a fixed value for each item i, we use a time-varying “proxy”\\\\nvalue that depends on the state of i’s Markov chain.\\\\n– Instead of immediately selecting the item i of greatest marginal value, we\\\\nadvance i’s Markov chain one step.\\\\nThe main difficulty lies in choosing each item’s proxy value, for which simple\\\\nheuristics can be suboptimal. We use a quantity for each state of each item’s\\\\nMarkov chain called its grade, and an item’s proxy value is its minimum grade so\\\\nfar. A state’s grade is closely related to the Gittins index from the multi-armed\\\\nbandit literature, which we discuss along with other related work in §6.\\\\n2 The Markovian Price of Information Model\\\\nTo capture the evolution of our knowledge about an item’s value, we use the\\\\nnotion of a Markov system from [17] (who did not consider values at the desti-\\\\nnations).\\\\nDefinition 2.1 (Markov System). A Markov system S = (V, P, s, T,pi, r) for\\\\nan element consists of a discrete Markov chain with state space V , a transition\\\\nmatrix P = {pu,v} indexed by V ×V (here pu,v is the probability of transitioning\\\\nfrom u to v), a starting state s, a set of absorbing destination states T ⊆ V ,\\\\na non-negative probing price πu ∈ R≥0 for every state u ∈ V \\\\\\\\ T , and a value\\\\nrt ∈ R for each destination state t ∈ T . We assume that every state u ∈ V\\\\nreaches some destination state.\\\\nWe have a collection J of ground elements, each associated with its own\\\\nMarkov system. An element is ready if its Markov system has reached one of its\\\\nabsorbing destination states. For a ready element, if ω is the (random) trajectory\\\\nof its Markov chain then d(ω) denotes its associated destination state. We now\\\\ndefine the Markovian PoI game, which consists of an objective function on J .\\\\nDefinition 2.2 (Markovian PoI Game). Given a set of ground elements J ,\\\\nconstraints F ⊆ 2J , an objective function f : 2J × R|J| → R, and a Markov\\\\nsystem Si = (Vi, Pi, si, Ti,pii, ri) for each element i ∈ J , the Markovian PoI\\\\ngame is the following. At each time step, we either advance a Markov system Si\\\\nfrom its current state u ∈ Vi \\\\\\\\ Ti by incurring price πui , or we end the game by\\\\nselecting a subset of ready elements I ⊆ J that are feasible—i.e., I ∈ F .\\\\nA common choice for f is the additive objective f(I,x) =\\\\n∑\\\\ni∈I xi.\\\\nLet ω denote the trajectory profile for the Markovian PoI game: it consists\\\\nof the random trajectories ωi taken by all the Markov chains i at the end of the\\\\ngame. To avoid confusion, we write the selected feasible solution I as I(ω). A\\\\nutility/disutility optimization problem is to give a strategy for a Markovian\\\\nPoI game while optimizing both the objective and the total price.\\\\n4 Gupta et al.\\\\nUtility Maximization (Util-Max): A Markovian PoI game where the\\\\nconstraints F are downward-closed (i.e., packing) and the values ri are non-\\\\nnegative for every i ∈ J (i.e., ∀t ∈ Ti, r\\\\nt\\\\ni ≥ 0, and can be understood as a reward\\\\nobtained for selecting i). The goal is to find a strategy ALG maximizing utility:\\\\nUmax(ALG)\\\\n∆\\\\n= Eω\\\\n[\\\\nf\\\\n(\\\\nI(ω), {r\\\\nd(ωi)\\\\ni }i∈I(ω)\\\\n)\\\\n︸ ︷︷ ︸\\\\nvalue\\\\n−\\\\n∑\\\\ni\\\\n∑\\\\nu∈ωi\\\\nπui︸ ︷︷ ︸\\\\ntotal price\\\\n]\\\\n. (1)\\\\nSince the empty set is always feasible, the optimum utility is non-negative.\\\\nWe also define a minimization variant of the problem that is useful to capture\\\\ncovering combinatorial problems such as minimum spanning trees and set cover.\\\\nDisutility Minimization (Disutil-Min) : A Markovian PoI game where\\\\nthe constraints F are upward-closed (i.e., covering) and the values ri are non-\\\\nnegative for every i ∈ J (i.e., ∀t ∈ Ti, rti ≥ 0, and can be understood as a cost\\\\nwe pay for selecting i). The goal is to find a strategy ALG minimizing disutility:\\\\nUmin(ALG)\\\\n∆\\\\n= Eω\\\\n[\\\\nf\\\\n(\\\\nI(ω), {r\\\\nd(ωi)\\\\ni }i∈I(ω)\\\\n)\\\\n+\\\\n∑\\\\ni\\\\n∑\\\\nu∈ωi\\\\nπui\\\\n]\\\\n.\\\\nWe will assume that the function f is non-negative when all ri are non-negative.\\\\nHence, the disutility of the optimal policy is non-negative.\\\\nIn the special case where all the Markov chains for a Markovian PoI game\\\\nare formed by a directed acyclic graph (Dag), we call the corresponding opti-\\\\nmization problem Dag-Util-Max or Dag-Disutil-Min.\\\\n3 Adaptive Utility Maximization via Frugal Algorithms\\\\nFrugal algorithms, introduced in Singla [37], capture the intuitive notion of\\\\n“greedy” algorithms. There are many known Frugal algorithms, e.g., optimal\\\\nalgorithms for matroids and O(1)-approx algorithms for matchings, vertex cover,\\\\nand facility location. These Frugal algorithms were designed in the traditional\\\\nfree information (Free-Info) setting, where each ground element has a fixed\\\\nvalue. Can we use them in the Markovian PoI world?\\\\nOur main contribution is a technique that adapts any Frugal algorithm to\\\\nthe Markovian PoI world, achieving the same approximation ratio as the orig-\\\\ninal algorithm. The result applies to semiadditive objective functions f , which\\\\nare those of the form f(I,x) =\\\\n∑\\\\ni∈I xi + h(I) for some h : 2\\\\nJ → R.\\\\nTheorem 3.1. For a semiadditive objective function val, if there exists an α-\\\\napproximation Frugal algorithm for a Util-Max problem over some packing\\\\nconstraints F in the Free-Info world, then there exists an α-approximation\\\\nstrategy for the corresponding Util-Max problem in the Markovian PoI world.\\\\nWe prove an analogous result for Disutil-Min in §D. The following corollar-\\\\nies immediately follow from known Frugal algorithms [37].\\\\nThe Markovian Price of Information 5\\\\nCorollary 3.1. In the Markovian PoI world, we have:\\\\n– An optimal algorithm for both Util-Max and Disutil-Min for matroids.\\\\n– A 2-approx for Util-Max for matchings and a k-approx for a k-system.\\\\n– A min{f, logn}-approx for Disutil-Min for set-cover, where f is the max-\\\\nimum number of sets in which a ground element is present.\\\\n– A 1.861-approx for Disutil-Min for facility location.\\\\n– A 3-approx for Disutil-Min for prize-collecting Steiner tree.\\\\nBefore proving Theorem 3.1, we define a grade for every state in a Markov\\\\nsystem in §3.1, much as in [17]. This grade is a variant of the popular Gittins\\\\nindex. In §3.2, we use the grade to define a prevailing cost and an epoch for\\\\na trajectory. In §3.3, we use these definitions to prove Theorem 3.1. We con-\\\\nsider Util-Max throughout, but analogous definitions and arguments hold for\\\\nDisutil-Min.\\\\n3.1 Grade of a State\\\\nTo define the grade τv of a state v ∈ V in Markov system S = (V, P, s, T,pi, r), we\\\\nconsider the following Markov game called τ-penalized S, denoted S(τ). Roughly,\\\\nS(τ) is the same as S but with a termination penalty, which is a constant τ ∈ R.\\\\nSuppose v ∈ V denotes the current state of S in the game S(τ). In each\\\\nmove, the player has two choices: (a) Halt that immediately ends the game, and\\\\n(b) Play that changes the state, price, and value as follows:\\\\n– If v ∈ V \\\\\\\\T , the player pays price πv, the current state of S changes according\\\\nto the transition matrix P , and the game continues.\\\\n– If v ∈ T , then the player receives penalized value rv − τ , where τ is the\\\\naforementioned termination penalty, and the game ends.\\\\nThe player wishes to maximize his utility, which is the expected value he\\\\nobtains minus the expected price he pays. We write Uv(τ) for the utility attained\\\\nby optimal play starting from state v ∈ V .\\\\nThe utility Uv(τ) is clearly non-increasing in the penalty τ , and one can also\\\\nshow that it is continuous [17, Section 4]. In the case of large penalty τ → +∞,\\\\nit is optimal to halt immediately, achieving Uv(τ) = 0. In the opposite extreme\\\\nτ → −∞, it is optimal to play until completion, achieving Uv(τ) → +∞. Thus,\\\\nas we increase τ from −∞ to +∞, the utility Uv(τ) becomes 0 at some critical\\\\nvalue τ = τv. This critical value τv that depends on state v is the grade.\\\\nDefinition 3.1 (Grade). The grade of a state v in Markov system S is τv\\\\n∆\\\\n=\\\\nsup{τ ∈ R | Uv(τ) > 0}. For a Util-Max problem, we write the grade of a state\\\\nv in Markov system Si corresponding to element i as τvi .\\\\nThe quantity grade of a state is well-defined from the above \\\\ndiscussion. We\\\\nemphasize that it is independent of all other Markov systems. Put another way,\\\\nthe grade of a state is the penalty τ that makes the player indifferent between\\\\nhalting and playing. It is known how to compute grade efficiently [17, Section 7].\\\\n6 Gupta et al.\\\\n3.2 Prevailing Cost and Epoch\\\\nWe now define a prevailing cost [17] and an epoch. The prevailing cost of Markov\\\\nsystem S is its minimum grade at any point in time.\\\\nDefinition 3.2 (Prevailing Cost). The prevailing cost of Markov system Si\\\\nin a trajectory ωi is Y\\\\nmax(ωi) = minv∈ωi{τ\\\\nv\\\\ni }. For trajectory profile ω, denote\\\\nY max(ω) the list of prevailing costs for each Markov system.\\\\nPut another way, the prevailing cost is the maximum termination penalty\\\\nfor the game S(τ) such that for every state along ω the player does not want to\\\\nhalt.\\\\nObserve that the prevailing cost of a trajectory can only decrease as it extends\\\\nfurther. In particular, it decreases whenever the Markov system reaches a state\\\\nwith grade smaller than each of the previously visited states. We can therefore\\\\nview the prevailing cost as a non-increasing piecewise constant function of time.\\\\nThis motivates us to define an epoch.\\\\nDefinition 3.3 (Epoch). An epoch for a trajectory ω is any maximal contin-\\\\nuous segment of ω where the prevailing cost does not change.\\\\nSince the grade can be computed efficiently, we can also compute the prevailing\\\\ncost and epochs of a trajectory efficiently.\\\\n3.3 Adaptive Algorithms for Utility Maximization\\\\nIn this section, we prove Theorem 3.1 that adapts a Frugal algorithm in Free-\\\\nInfo world to a probing strategy in the Markovian PoI world. This theorem\\\\nconcerns semiadditive functions, which are useful to capture non-additive objec-\\\\ntives of problems like facility location and prize-collecting Steiner tree.\\\\nDefinition 3.4 (Semiadditive Function [37]). A function f(I,X) : 2J ×\\\\nR|J| → R is semiadditive if there exists a function h : 2J → R s.t. f(I,x) =∑\\\\ni∈I xi + h(I).\\\\nAll additive functions are semiadditive with h(I) = 0 for all I. To capture the\\\\nfacility location problem on a graph G = (J,E) with metric (J, d), clients C ⊆ J ,\\\\nand facility opening costs x : J → R≥0, we can define h(I) =\\\\n∑\\\\nj∈C mini∈I d(j, i).\\\\nNotice h only depends on the identity of facilities I and not their opening costs.\\\\nThe proof of Theorem 3.1 takes two steps. We first give a randomized re-\\\\nduction to upper bound the utility of the optimal strategy in the Markovian\\\\nPoI world with the optimum of a surrogate problem in the Free-Info world.\\\\nThen, we transform a Frugal algorithm into a strategy with utility close to\\\\nthis bound.\\\\nThe Markovian Price of Information 7\\\\nUpper Bounding the Optimal Strategy Using a Surrogate. The main\\\\nidea in this section is to show that for Util-Max, no strategy (in particular,\\\\noptimal) can derive more utility from an element i ∈ J than its prevailing cost.\\\\nHere, the prevailing cost of i is for a random trajectory to a destination state\\\\nin Markov system Si. Since the optimal strategy can only select a feasible set\\\\nin F , this idea naturally leads to the following Free-Info surrogate problem:\\\\nimagine each element’s value is exactly its (random) prevailing cost, the goal is\\\\nto select a set feasible in F to maximize the total value. In Lemma 3.1, we show\\\\nthat the expected optimum value of this surrogate problem is an upper bound\\\\non the optimum utility for Util-Max. First, we formally define the surrogate\\\\nproblem.\\\\nDefinition 3.5 (Surrogate Problem). Given a Util-Max problem with semi-\\\\nadditive objective val and packing constraints F over universe J , the correspond-\\\\ning surrogate problem over J is the following. It consists of constraints F and\\\\n(random) objective function f˜ : 2J → R given by f˜(I) = val(I,Ymax(ω)), where\\\\nYmax(ω) denotes the prevailing costs over a random trajectory profile ω consist-\\\\ning of independent random trajectories for each element i ∈ J to a destination\\\\nstate. The goal is to select I ∈ F to maximize f˜(I).\\\\nLet SUR(ω)\\\\n∆\\\\n= maxI∈F{val(I,Ymax(ω))} denote the o"},"Similar Full Paper":{"kind":"string","value":"{'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\\\nABSTRACT\\\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\\\nquality of visual generative models. However, in autoregressive (AR) multi-modal\\\\ngeneration, CFG introduces design inconsistencies between language and visual\\\\ncontent, contradicting the design philosophy of unifying different modalities for\\\\nvisual AR. Motivated by language model alignment methods, we propose Condi-\\\\ntion Contrastive Alignment (CCA) to facilitate guidance-free AR visual generation\\\\nwith high performance and analyze its theoretical connection with guided sampling\\\\nmethods. Unlike guidance methods that alter the sampling process to achieve the\\\\nideal sampling distribution, CCA directly fine-tunes pretrained models to fit the\\\\nsame distribution target. Experimental results show that CCA can significantly\\\\nenhance the guidance-free performance of all tested models with just one epoch\\\\nof fine-tuning ( ∼1% of pretraining epochs) on the pretraining dataset, on par with\\\\nguided sampling methods. This largely removes the need for guided sampling in\\\\nAR visual generation and cuts the sampling cost by half. Moreover, by adjusting\\\\ntraining parameters, CCA can achieve trade-offs between sample diversity and\\\\nfidelity similar to CFG. This experimentally confirms the strong theoretical connec-\\\\ntion between language-targeted alignment and visual-targeted guidance methods,\\\\nunifying two previously independent research fields. Code and model weights:\\\\nhttps://github.com/thu-ml/CCA .\\\\n/uni00000014/uni00000014/uni00000014/uni00000030 /uni00000016/uni00000017/uni00000016/uni00000030 /uni0000001a/uni0000001a/uni00000018/uni00000030 /uni00000014/uni00000011/uni00000017/uni00000025 /uni00000016/uni00000011/uni00000014/uni00000025/uni00000013/uni00000018/uni00000014/uni00000013/uni00000014/uni00000018/uni00000015/uni00000013/uni00000015/uni00000018/uni00000016/uni00000013/uni00000029/uni0000002c/uni00000027/uni00000003\\\\n/uni0000001a/uni00000011/uni00000013/uni00000017\\\\n/uni00000016/uni00000011/uni00000017/uni00000016 /uni00000016/uni00000011/uni00000014 /uni00000016/uni00000011/uni00000014/uni00000015 /uni00000015/uni00000011/uni00000019/uni0000001c/uni00000016/uni00000015/uni00000011/uni00000015/uni00000013\\\\n/uni00000014/uni0000001c/uni00000011/uni00000013/uni0000001a\\\\n/uni00000014/uni00000018/uni00000011/uni00000018/uni00000018/uni00000014/uni00000017/uni00000011/uni00000019/uni00000018\\\\n/uni0000001c/uni00000011/uni00000016/uni0000001b\\\\n/uni00000019/uni00000011/uni00000013/uni0000001c\\\\n/uni00000016/uni00000011/uni00000013/uni0000001b/uni00000015/uni00000011/uni00000019/uni00000015 /uni00000015/uni00000011/uni00000016/uni00000017 /uni00000015/uni00000011/uni00000014/uni0000001b/uni00000026/uni00000026/uni00000024/uni00000003/uni0000000b/uni00000052/uni00000058/uni00000055/uni00000056/uni0000000c\\\\n/uni0000005a/uni00000012/uni00000052/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n/uni0000005a/uni00000012/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n/uni00000014/uni00000014/uni00000014/uni00000030 /uni00000016/uni00000017/uni00000016/uni00000030 /uni0000001a/uni0000001a/uni00000018/uni00000030 /uni00000014/uni00000011/uni00000017/uni00000025 /uni00000016/uni00000011/uni00000014/uni00000025/uni00000018/uni00000013/uni00000014/uni00000013/uni00000013/uni00000014/uni00000018/uni00000013/uni00000015/uni00000013/uni00000013/uni00000015/uni00000018/uni00000013/uni00000016/uni00000013/uni00000013/uni00000016/uni00000018/uni00000013/uni0000002c/uni00000036/uni00000003\\\\n/uni00000015/uni00000015/uni00000013/uni00000011/uni0000001a/uni00000015/uni0000001b/uni0000001b/uni00000011/uni00000015/uni00000016/uni00000016/uni0000001c/uni00000011/uni00000013/uni00000016/uni00000015/uni0000001c/uni00000011/uni0000001b\\\\n/uni00000015/uni0000001a/uni00000019/uni00000011/uni0000001b\\\\n/uni00000016/uni0000001c/uni00000011/uni0000001c/uni00000019/uni00000017/uni00000011/uni00000016/uni0000001a/uni0000001c/uni00000011/uni00000015/uni0000001b/uni00000019/uni00000011/uni00000016/uni00000014/uni00000014/uni00000015/uni00000011/uni0000001c/uni00000014/uni0000001b/uni00000015/uni00000011/uni00000018/uni00000015/uni00000018/uni00000019/uni00000011/uni00000014/uni00000015/uni00000017/uni00000017/uni00000011/uni00000014\\\\n/uni00000015/uni00000018/uni00000016/uni00000011/uni0000001c/uni00000015/uni00000019/uni00000016/uni00000011/uni00000016\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni0000000b/uni00000052/uni00000058/uni00000055/uni00000056/uni0000000c\\\\n/uni0000005a/uni00000012/uni00000052/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n/uni0000005a/uni00000012/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n(a) LlamaGen\\\\n/uni00000016/uni00000014/uni00000013/uni00000030 /uni00000019/uni00000013/uni00000013/uni00000030 /uni00000014/uni00000011/uni00000013/uni00000025 /uni00000015/uni00000011/uni00000013/uni00000025/uni00000015/uni00000017/uni00000019/uni0000001b/uni00000014/uni00000013/uni00000014/uni00000015/uni00000029/uni0000002c/uni00000027/uni00000003\\\\n/uni00000017/uni00000011/uni00000013/uni00000016\\\\n/uni00000016/uni00000011/uni00000013/uni00000015\\\\n/uni00000015/uni00000011/uni00000019/uni00000016 /uni00000015/uni00000011/uni00000018/uni00000017/uni00000014/uni00000015/uni00000011/uni00000013/uni00000013\\\\n/uni0000001b/uni00000011/uni00000017/uni0000001b\\\\n/uni00000019/uni00000011/uni00000015/uni00000013\\\\n/uni00000018/uni00000011/uni00000015/uni00000019\\\\n/uni00000016/uni00000011/uni00000016/uni00000019\\\\n/uni00000015/uni00000011/uni00000018/uni0000001c\\\\n/uni00000015/uni00000011/uni00000014/uni00000014 /uni00000014/uni00000011/uni0000001c/uni00000017/uni00000026/uni00000026/uni00000024/uni00000003/uni0000000b/uni00000052/uni00000058/uni00000055/uni00000056/uni0000000c\\\\n/uni0000005a/uni00000012/uni00000052/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n/uni0000005a/uni00000012/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n/uni00000016/uni00000014/uni00000013/uni00000030 /uni00000019/uni00000013/uni00000013/uni00000030 /uni00000014/uni00000011/uni00000013/uni00000025 /uni00000015/uni00000011/uni00000013/uni00000025/uni00000014/uni00000013/uni00000013/uni00000014/uni00000018/uni00000013/uni00000015/uni00000013/uni00000013/uni00000015/uni00000018/uni00000013/uni00000016/uni00000013/uni00000013/uni0000002c/uni00000036/uni00000003\\\\n/uni00000015/uni0000001a/uni00000014/uni00000011/uni00000014/uni00000015/uni0000001b/uni00000015/uni00000011/uni00000019/uni00000015/uni0000001c/uni0000001b/uni00000011/uni0000001b\\\\n/uni00000015/uni00000019/uni00000017/uni00000011/uni00000015\\\\n/uni00000014/uni00000013/uni00000018/uni00000011/uni00000018/uni00000014/uni00000015/uni0000001c/uni00000011/uni00000018/uni00000014/uni00000018/uni00000017/uni00000011/uni00000016/uni00000014/uni0000001a/uni00000018/uni00000011/uni00000019/uni00000015/uni0000001b/uni00000013/uni00000011/uni00000014\\\\n/uni00000015/uni0000001a/uni0000001c/uni00000011/uni00000014/uni00000015/uni0000001c/uni00000014/uni00000011/uni00000019/uni00000016/uni00000015/uni00000018/uni00000011/uni00000014\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni0000000b/uni00000052/uni00000058/uni00000055/uni00000056/uni0000000c\\\\n/uni0000005a/uni00000012/uni00000052/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048\\\\n/uni0000005a/uni00000012/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048 (b) V AR\\\\nFigure 1: CCA significantly improves guidance-free sample quality for AR visual generative models\\\\nwith just one epoch of fine-tuning on the pretraining dataset.\\\\n1 I NTRODUCTION\\\\nWitnessing the scalability and generalizability of autoregressive (AR) models in language domains,\\\\nrecent works have been striving to replicate similar success for visual generation (Esser et al., 2021;\\\\nLee et al., 2022). By quantizing images into discrete tokens, AR visual models can process images\\\\nusing the same next-token prediction approach as Large Language Models (LLMs). This approach is\\\\nattractive because it provides a potentially unified framework for vision and language, promoting\\\\nconsistency in reasoning and generation across modalities (Team, 2024; Xie et al., 2024).\\\\nDespite the design philosophy of maximally aligning visual modeling with language modeling\\\\nmethods, AR visual generation still differs from language generation in a notable aspect. AR visual\\\\n1arXiv:2410.09347v1 [cs.CV] 12 Oct 2024Preprint\\\\ngeneration relies heavily on Classifier-Free Guidance (CFG) (Ho & Salimans, 2022), a sampling\\\\ntechnique unnecessary for language generation, which has caused design inconsistencies between\\\\nthe two types of content. During sampling, while CFG helps improve sample quality by contrasting\\\\nconditional and unconditional models, it requires two model inferences per visual token, which\\\\ndoubles the sampling cost. During training, CFG requires randomly masking text conditions to learn\\\\nthe unconditional distribution, preventing the simultaneous training of text tokens (Team, 2024).\\\\nIn contrast to visual generation, LLMs rarely rely on guided sampling. Instead, the surge of LLMs’\\\\ninstruction-following abilities has largely benefited from fine-tuning-based alignment methods (Schul-\\\\nman et al., 2022). Motivated by this observation, we seek to study: “Can we avoid guided sampling\\\\nin AR visual generation, but attain similar effects by directly fine-tuning pretrained models?”\\\\nIn this paper, we derive Condition Contrastive Alignment (CCA) for enhancing visual AR performance\\\\nwithout guided sampling. Unlike CFG which necessitates altering the sampling process to achieve a\\\\nmore desirable sampling distribution, CCA directly fine-tunes pretrained AR models to fit the same\\\\ndistribution target, leaving the sampling scheme untouched. CCA is quite convenient to use since\\\\nit does not rely on any additional datasets beyond the pretraining data. Our method functions by\\\\ncontrasting positive and negative conditions for a given image, which can be easily created from the\\\\nexisting pretraining dataset as matched or mismatched image-condition pairs. CCA is also highly\\\\nefficient given its fine-tuning nature. We observe that our method achieves ideal performance within\\\\njust one training epoch, indicating negligible computational overhead ( ∼1% of pretraining).\\\\nIn Sec. 4, we highlight a theoretical connection between CCA and guided sampling techniques\\\\n(Dhariwal & Nichol, 2021; Ho & Salimans, 2022). Essentially these methods all target at the same\\\\nsampling distribution. The distributional gap between this target distribution and pretrained models is\\\\nrelated to a physical quantity termed conditional residual ( logp(x|c)\\\\np(x)). Guidance methods typically\\\\ntrain an additional model (e.g., unconditional model or classifier) to estimate this quantity and enhance\\\\npretrained models by altering their sampling process. Contrastively, CCA follows LLM alignment\\\\ntechniques (Rafailov et al., 2023; Chen et al., 2024a) and parameterizes the conditional residual\\\\nwith the difference between our target model and the pretrained model, thereby directly training a\\\\nsampling model. This analysis unifies language-targeted alignment and visual-targeted guidance\\\\nmethods, bridging the gap between the two previously independent research fields.\\\\nWe apply CCA to two state-of-the-art autoregressive (AR) visual models, LLamaGen (Sun et al.,\\\\n2024) and V AR (Tian et al., 2024), which feature distinctly different visual tokenization designs.\\\\nBoth quantitative and qualitative results show that CCA significantly and consistently enhances the\\\\nguidance-free sampling quality across all tested models, achieving performance levels comparable\\\\nto CFG (Figure 1). We further show that by varying training hyperparameters, CCA can realize a\\\\ncontrollable trade-off between image diversity and fidelity similar to CFG. This further confirms their\\\\ntheoretical connections. We also compare our method with some existing LLM alignment methods\\\\n(Welleck et al., 2019; Rafailov et al., 2023) to justify its algorithm design. Finally, we demonstrate\\\\nthat CCA can be combined with CFG to further improve performance.\\\\nOur contributions: 1. We take a big step toward guidance-free visual generation by significantly\\\\nimproving the visual quality of AR models. 2. We reveal a theoretical connection between alignment\\\\nand guidance methods. This shows that language-targeted alignment can be similarly applied to\\\\nvisual generation and effectively replace guided sampling, closing the gap between these two fields.\\\\n2 B ACKGROUND\\\\n2.1 A UTOREGRESSIVE (AR) V ISUAL MODELS\\\\nAutoregressive models. Consider data xrepresented by a sequence of discrete tokens x1:N:=\\\\n{x1, x2, ..., x N}, where each token xnis an integer. Data probability p(x)can be decomposed as:\\\\np(x) =p(x1)NY\\\\nn=2p(xn|x<n). (1)\\\\nAR models thus aim to learn pϕ(xn|x<n)≈p(xn|x<n), where each token xnis conditioned only\\\\non its previous input x<n. This is known as next-token prediction (Radford et al., 2018).\\\\n2Preprint\\\\nVisual tokenization. Image pixels are continuous values, making it necessary to use vector-\\\\nquantized tokenizers for applying discrete AR models to visual data (Van Den Oord et al., 2017;\\\\nEsser et al., 2021). These tokenizers are trained to encode images xinto discrete token sequences\\\\nx1:Nand decode them back by minimizing reconstruction losses. In our work, we utilize pretrained\\\\nand frozen visual tokenizers, allowing AR models to process images similarly to text.\\\\n2.2 G UIDED SAMPLING FOR VISUAL GENERATION\\\\nDespite the core motivation of developing a unified model for language and vision, the AR sampling\\\\nstrategies for visual and text contents differ in one key aspect: AR visual generation necessitates a\\\\nsampling technique named Classifier-Free Guidance (CFG) (Ho & Salimans, 2022). During inference,\\\\nCFG adjusts the sampling logits ℓsamplefor each token as:\\\\nℓsample=ℓc+s(ℓc−ℓu), (2)\\\\nwhere ℓcandℓuare the conditional and unconditional logits provided by two separate AR models,\\\\npϕ(x|c)andpϕ(x). The condition ccan be class labels or text captions, formalized as prompt tokens.\\\\nThe scalar sis termed guidance scale. Since token logits represent the (unnormalized) log-likelihood\\\\nin AR models, Ho & Salimans (2022) prove that the sampling distribution satisfies:\\\\npsample(x|c)∝pϕ(x|c)\\\\x14pϕ(x|c)\\\\npϕ(x)\\\\x15s\\\\n. (3)\\\\nAts= 0, the sampling model becomes exactly the pretrained conditional model pϕ. However,\\\\nprevious works (Ho & Salimans, 2022; Podell et al., 2023; Chang et al., 2023; Sun et al., 2024) have\\\\nwidely observed that an appropriate s >0is critical for an ideal trade-off between visual fidelity and\\\\ndiversity, making training another unconditional model pϕnecessary. In practice, the unconditional\\\\nmodel usually shares parameters with the conditional one, and can be trained concurrently by\\\\nrandomly dropping condition prompts cduring training.\\\\nOther guidance methods, such as Classifier Guidance (Ho & Salimans, 2022) and Energy Guidance\\\\n(Lu et al., 2023) have similar effects of CFG. The target sampling distribution of these methods can\\\\nall be unified under Eq. 3.\\\\n2.3 D IRECT PREFERENCE OPTIMIZATION FOR LANGUAGE MODEL ALIGNMENT\\\\nReinforcement Learning from Human Feedback (RLHF) is crucial for enhancing the instruction-\\\\nfollowing ability of pretrained Language Models (LMs) (Schulman et al., 2022; OpenAI, 2023).\\\\nPerforming RL typically requires a reward model, which can be learned from human preference data.\\\\nFormally, the Bradley-Terry preference model (Bradley & Terry, 1952) assumes.\\\\np(xw≻xl|c) :=er(c,xw)\\\\ner(c,xl)+er(c,xw)=σ(r(c,xw)−r(c,xl)), (4)\\\\nwhere xwandxlare respectively the winning and losing response for an instruction c, evaluated\\\\nby human. r(·)represents an implicit reward for each response. The target LM πθshould satisfy\\\\nπθ(x|c)∝µϕ(x|c)er(c,x)/βto attain higher implicit reward compared with the pretrained LM µϕ.\\\\nDirect Preference Optimization (Rafailov et al., 2023) allows us to directly optimize pretrained LMs\\\\non preference data, by formalizing rθ(c,x) :=βlogπθ(x|c)−βlogµϕ(x|c):\\\\nLDPO\\\\nθ=−E{c,xw≻xl}logσ\\\\x12\\\\nβlogπθ(xw|c)\\\\nµϕ(xw|c)−βlogπθ(xl|c)\\\\nµϕ(xl|c)\\\\x13\\\\n. (5)\\\\nDPO is more streamlined and thus often more favorable compared with traditional two-stage RLHF\\\\npipelines: first training reward models, then aligning LMs with reward models using RL.\\\\n3 C ONDITION CONTRASTIVE ALIGNMENT\\\\nAutoregressive visual models are essentially learning a parameterized model pϕ(x|c)to approximate\\\\nthe standard conditional image distribution p(x|c). Guidance algorithms shift the sampling policy\\\\n3Preprint\\\\npsample(x|c)away from p(x|c)according to Sec. 2.2:\\\\npsample(x|c)∝p(x|c)\\\\x14p(x|c)\\\\np(x)\\\\x15s\\\\n. (6)\\\\nAt guidance scale s= 0, sampling from psample(x|c) =p(x|c)≈pϕ(x|c)is most straightforward.\\\\nHowever, it is widely observed that an appropriate s >0usually leads to significantly enhanced\\\\nsample quality. The cost is that we rely on an extra unconditional model pϕ(x)≈p(x)for sampling.\\\\nThis doubles the sampling cost and causes an inconsistent training paradigm with language.\\\\nIn this section, we derive a simple approach to directly model the same target distribution psample\\\\nby fine-tuning pretrained models. Specifically, our methods leverage a singular loss function for\\\\noptimizing pϕ(x|c)≈p(x|c)to become psample\\\\nθ(x|c)≈psample(x|c). Despite having similar effects\\\\nas guided sampling, our approach does not require altering the sampling process. We theoretically\\\\nderive our method in Sec. 3.1 and discuss its practical implementation in Sec. 3.2.\\\\n3.1 A LGORITHM DERIVATION\\\\nThe core difficulty of directly learning psample\\\\nθis that we cannot access datasets under the distribution\\\\nofpsample. However, we observe the distributional difference between psample(x|c)andp(x|c)is\\\\nrelated to a simple quantity that can be potentially learned from existing datasets. Specifically, by\\\\ntaking the logarithm of both sides in Eq. 6 and applying some algebra, we have1:\\\\n1\\\\nslogpsample(x|c)\\\\np(x|c)= logp(x|c)\\\\np(x), (7)\\\\nof which the right-hand side (i.e., logp(x|c)\\\\np(x)) corresponds to the log gap between the conditional\\\\nprobability and unconditional probability for an image x, which we term as conditional residual .\\\\nOur key insight here is that the conditional residual can be directly learned through contrastive\\\\nlearning approaches (Gutmann & Hyvärinen, 2012), as sated below:\\\\nTheorem 3.1 (Noise Contrastive Estimation, proof in Appendix A) .Letrθbe a parameterized model\\\\nwhich takes in an image-condition pair (x,c)and outputs a scalar value rθ(x,c). Consider the loss\\\\nfunction:\\\\nLNCE\\\\nθ(x,c) =−Ep(x,c)logσ(rθ(x,c))−Ep(x)p(c)logσ(−rθ(x,c)). (8)\\\\nGiven unlimited model expressivity for rθ, the optimal solution for minimizing LNCE\\\\nθsatisfies\\\\nr∗\\\\nθ(x,c) = logp(x|c)\\\\np(x). (9)\\\\nNow that we have a tractable way of learning rθ(x,c)≈logp(x|c)\\\\np(x), the target distribution psamplecan\\\\nbe jointly defined by rθ(x,c)and the pretrained model pϕ. However, we would still lack an explicitly\\\\nparameterized model psample\\\\nθifrθ(x,c)is another independent network. To address this problem, we\\\\ndraw inspiration from the widely studied alignment techniques in language models (Rafailov et al.,\\\\n2023) and parameterize rθ(x,c)with our target model psample\\\\nθ(x|c)andpϕ(x|c)according to Eq. 7:\\\\nrθ(x,c) :=1\\\\nslogpsample\\\\nθ(x|c)\\\\npϕ(x|c). (10)\\\\nThen, the loss function becomes\\\\nLCCA\\\\nθ=−Ep(x,c)logσh1\\\\nslogpsample\\\\nθ(x|c)\\\\npϕ(x|c)i\\\\n−Ep(x)p(c)logσh\\\\n−1\\\\nslogpsample\\\\nθ(x|c)\\\\npϕ(x|c)i\\\\n. (11)\\\\nDuring training, psample\\\\nθis learnable while pretrained pϕis frozen. psample\\\\nθcan be initialized from pϕ.\\\\nThis way we can fit psamplewith a single AR model psample\\\\nθ, eliminating the need for training a separate\\\\nunconditional model for guided sampling. Sampling strategies for psample\\\\nθare consistent with standard\\\\nlanguage model decoding methods, which unifies decoding systems for multi-modal generation.\\\\n1We ignore a normalizing constant in Eq. 7 for brevity. A more detailed discussion is in Appendix B.\\\\n4Preprint\\\\n…\\\\n{𝑥2,𝑐1}\\\\n{𝑥3,𝑐1}\\\\n{𝑥𝐾,𝑐1}{𝑥1,𝑐2}\\\\n{𝑥3,𝑐2}\\\\n{𝑥𝐾,𝑐2}{𝑥1,𝑐3}\\\\n{𝑥2,𝑐3}\\\\n{𝑥𝐾,𝑐3}{𝑥1,𝑐𝐾}\\\\n{𝑥2,𝑐𝐾}\\\\n{𝑥3,𝑐𝐾}\\\\n{𝑥𝐾,𝑐𝐾}\\\\n𝑥1\\\\n𝑥2\\\\n𝑥3\\\\n𝑥𝐾…𝑐1\\\\n<Cat>𝑐2\\\\n<Dog>𝑐3\\\\n<Bird>𝑐𝐾\\\\n<Van>\\\\n{𝑥1,𝑐1}\\\\n{𝑥2,𝑐2}\\\\n{𝑥3,𝑐3}\\\\n …\\\\nmax\\\\n𝜃log𝜎log𝑝𝜃ȁ𝑥𝑐\\\\n𝑝𝜙ȁ𝑥𝑐\\\\nlog𝜎log𝑝𝜃\\\\n𝑝𝜙\\\\nlog𝜎−log𝑝𝜃\\\\n𝑝𝜙s\\\\nmax\\\\n𝜃init{𝑥1,𝑐1}{𝑥1,𝑐1}{𝑥1,𝑐1}\\\\n 𝑥1\\\\n𝑥2\\\\n𝑥3\\\\n𝑥𝑁…𝑐1\\\\n<Cat>𝑐2\\\\n<Dog>𝑐3\\\\n<Bird>𝑐𝑁\\\\n<Van>\\\\n{𝑥1,𝑐1}\\\\n……………\\\\n…\\\\n…\\\\n…\\\\n(b) AR model likelihood\\\\nNegative data\\\\nPositive data\\\\n(a) Training batch (c) Alignment loss…\\\\n𝑝𝑥\\\\n𝑝𝑐𝑝𝑥,𝑐\\\\n𝑝𝑥𝑝𝑐\\\\nFigure 2: An overview of the CCA method.\\\\n3.2 P RACTICAL ALGORITHM\\\\nFigure 2 illustrates the CCA method. Specifically, implementing Eq. 11 requires approximating\\\\ntwo expectations: one under the joint distribution p(x,c)and the other under the product of its two\\\\nmarginals p(x)p(c). The key difference between these distributions is that in p(x,c), images xand\\\\nconditions care correctly paired. In contrast, xandcare sampled independently from p(x)p(c),\\\\nmeaning they are most likely mismatched.\\\\nIn practice, we rely solely on the pretraining dataset to estimate LCCA\\\\nθ. Consider a batch of Kdata\\\\npairs{x,c}1:K. We randomly shuffle the condition batch c1:Kto become cneg\\\\n1:K, where each cneg\\\\nkrepresents a negative condition of image xk, while the original ckis a positive one. This results in\\\\nour training batch {x,c,cneg}1:K. The loss function is\\\\nLCCA\\\\nθ(xk,ck,cneg\\\\nk) =−logσh\\\\nβlogpsample\\\\nθ(xk|ck)\\\\npϕ(xk|ck)i\\\\n| {z }\\\\nrelative likelihood for positive conditions ↑−λlogσh\\\\n−βlogpsample\\\\nθ(xk|cneg\\\\nk)\\\\npϕ(xk|cneg\\\\nk)i\\\\n| {z }\\\\nrelative likelihood for negative conditions ↓,(12)\\\\nwhere βandλare two hyperparameters that can be adjusted. βreplaces the guidance scale parameter\\\\ns, while λis for controlling the loss weight assigned to negative conditions. The learnable psample\\\\nθis\\\\ninitialized from the pretrained conditional model pϕ, making LCCA\\\\nθa fine-tuning loss.\\\\nWe give an intuitive understanding of Eq. 12. Note that logσ(·)is monotonically increasing. The\\\\nfirst term of Eq. 12 aims to increase the likelihood of an image xgiven a positive condition, with a\\\\nsimilar effect to maximum likelihood training. For mismatched image-condition data, the second\\\\nterm explicitly minimizes its relative model likelihood compared with the pretrained pϕ.\\\\nWe name the above training technique Condition Contrastive Alignment (CCA) due to its contrastive\\\\nnature in comparing positive and negative conditions with respect to a single image. This naming\\\\nalso reflects its theoretical connection with Noise Contrastive Estimation (Theorem 3.1).\\\\n4 C ONNECTION BETWEEN CCA AND GUIDANCE METHODS\\\\nAs summarized in Table 1, the key distinction between CCA and guidance methods is how to model\\\\nlogp(x|c)\\\\np(x), which defines the distributional gap between the target psample(x|c)andp(x|c)(Eq. 7).\\\\nIn particular, Classifier Guidance (Dhariwal & Nichol, 2021) leverages Bayes’ Rule and turn\\\\nlogp(x|c)\\\\np(x)into a conditional posterior:\\\\nlogp(x|c)\\\\np(x)= log p(c|x)−logp(c)≈logpθ(c|x)−logp(c),\\\\nwhere p(c|x)is explicitly modeled by a classifier pθ(c|x), which is trained by a standard classification\\\\nloss.p(c)is regarded as a uniform distribution. CFG trains an extra unconditional model pθ(x)to\\\\n5Preprint\\\\nMethod Classifier Guidance Classifier-Free Guidance Condition Contrastive Alignment\\\\nModeling of logp(x|c)\\\\np(x)logpθ(c|x)−logp(c) logpϕ(x|c)−logpθ(x) β[logpsample\\\\nθ(x|c)−logpϕ(x|c)]\\\\nTraining loss max θEp(x,c)logpθ(c|x) max θEp(x)logpθ(x) minθLCCA\\\\nθin Eq. 11\\\\nSampling policy logpϕ(x|c) +slogpθ(c|x)(1 +s) logpϕ(x|c)−slogpθ(x) logpsample\\\\nθ(x|c)\\\\nExtra training cost ∼9% of learning pϕ ∼10% of learning pϕ ∼1% of pretraining pϕ\\\\nSampling cost × ∼1.3 ×2 ×1\\\\nApplicable area Diffusion Diffusion & Autoregressive Autoregressive\\\\nTable 1: Comparison of CCA (ours) and guidance methods in visual generative models. Computa-\\\\ntional costs are estimated according to Dhariwal & Nichol (2021) and Ho & Salimans (2022).\\\\nestimate the unknown part of logp(x|c)\\\\np(x):\\\\nlogp(x|c)\\\\np(x)≈logpϕ(x|c)−logpθ(x).\\\\nDespite their effectiveness, guidance methods all require learning a separate model and a modified\\\\nsampling process compared with standard autoregressive decoding. In comparison, CCA leverages\\\\nEq. 7 and models logp(x|c)\\\\np(x)as\\\\nlogp(x|c)\\\\np(x)≈β[logpsample\\\\nθ(x|c)−logpϕ(x|c)],\\\\nwhich allows us to directly learn psample\\\\nθinstead of another guidance network.\\\\nAlthough CCA and conventional guidance techniques have distinct modeling methods, they all target\\\\nat the same sampling distribution and thus have similar effects in visual generation. For instance, we\\\\nshow in Sec. 5.2 that CCA offers a similar trade-off between sample diversity and fidelity to CFG.\\\\n5 E XPERIMENTS\\\\nWe seek to answer the following questions through our experiments:\\\\n1.How effective is CCA in enhancing the guidance-free generation quality of pretrained AR\\\\nvisual models, quantitatively and qualitatively? (Sec. 5.1)\\\\n2.Does CCA allow controllable trade-offs between sample diversity (FID) and fidelity (IS)\\\\nsimilar to CFG? (Sec. 5.2)\\\\n3. How does CCA perform in comparison to alignment algorithms for LLMs? (Sec. 5.3)\\\\n4. Can CCA be combined with CFG to further improve performance? (Sec. 5.4)\\\\n5.1 T OWARD GUIDANCE -FREE AR V ISUAL GENERATION\\\\nBase model. We experiment on two families of publicly accessible AR visual models, LlamaGen\\\\n(Sun et al., 2024) and V AR (Tian et al., 2024). Though both are class-conditioned models pretrained\\\\non ImageNet, LlamaGen and V AR feature distinctively different tokenizer and architecture designs.\\\\nLlamaGen focuses on reducing the inductive biases on visual signals. It tokenizes images in the\\\\nclassic raster order and adopts almost the same LLM architecture as Llama (Touvron et al., 2023a).\\\\nV AR, on the other hand, leverages the hierarchical structure of images and tokenizes them in a multi-\\\\nscale, coarse-to-fine manner. V AR adopts a GPT-2 architecture but tailors the attention mechanism\\\\nspecifically for visual content. For both works, CFG is a default and critical technique.\\\\nTraining setup. We leverage CCA to finetune multiple LlamaGen and V AR models of various sizes\\\\non the standard ImageNet dataset. The training scheme and hyperparameters are mostly consistent\\\\nwith the pretraining phase. We report performance numbers after only one training epoch and find\\\\nthis to be sufficient for ideal performance. We fix β= 0.02in Eq. 12 and select suitable λfor\\\\neach model. Image resolutions are 384×384for LlamaGen and 256×256for V AR. Following the\\\\noriginal work, we resize LlamaGen samples to 256×256whenever required for evaluation.\\\\n6Preprint\\\\nModelw/o Guidance w/ Guidance\\\\nFID↓ IS↑ Precision ↑Recall↑FID↓ IS↑DiffusionADM (Dhariwal & Nichol, 2021) 7.49 127.5 0.72 0.63 3.94 215.8\\\\nLDM-4 (Rombach et al., 2022) 10.56 103.5 0.71 0.62 3.60 247.7\\\\nU-ViT-H/2 (Bao et al., 2023) – – – – 2.29 263.9\\\\nDiT-XL/2 (Peebles & Xie, 2023) 9.62 121.5 0.67 0.67 2.27 278.2\\\\nMDTv2-XL/2 (Gao et al., 2023) 5.06 155.6 0.72 0.66 1.58 314.7MaskMaskGIT (Chang et al., 2022) 6.18 182.1 0.80 0.51 – –\\\\nMAGVIT-v2 (Yu et al., 2023) 3.65 200.5 – – 1.78 319.4\\\\nMAGE (Li et al., 2023) 6.93 195.8 – – – –AutoregressiveVQGAN (Esser et al., 2021) 15.78 74.3 – – 5.20 280.3\\\\nViT-VQGAN (Yu et al., 2021) 4.17 175.1 – – 3.04 227.4\\\\nRQ-Transformer (Lee et al., 2022) 7.55 134.0 – – 3.80 323.7\\\\nLlamaGen-3B (Sun et al., 2024) 9.38 112.9 0.69 0.67 2.18 263.3\\\\n+CCA (Ours) 2.69 276.8 0.80 0.59 – –\\\\nV AR-d30 (Tian et al., 2024) 5.25 175.6 0.75 0.62 1.92 323.1\\\\n+CCA (Ours) 2.54 264.2 0.83 0.56 – –\\\\nTable 2: Model comparisons on class-conditional ImageNet 256×256benchmark.\\\\nLlamaGen (w/o Guidance) LlamaGen + CCA (w/o G.) LlamaGen (w/ CFG)\\\\nIS=64.7 IS=384.6 IS=404.0\\\\nV AR (w/o Guidance) V AR + CCA (w/o G.) V AR (w/ CFGv2)\\\\nIS=154.3 IS=350.4 IS=390.8\\\\nFigure 3: CCA and CFG can similarly enhance the sample fidelity of AR visual models. The base\\\\nmodels are LlamaGen-L (343M) and V AR-d24 (1.0B). We use s= 3.0for CFG, and β= 0.02, λ=\\\\n104for CCA. Figure 7 and Figure 8 contain more examples.\\\\nExperimental results. We find CCA significantly improves the guidance-free performance of all\\\\ntested models (Figure 1), evaluated by metrics like FID (Heusel et al., 2017) and IS (Salimans et al.,\\\\n2016). For instance, after one epoch of CCA fine-tuning, the FID score of LlamaGen-L (343M)\\\\nimproves from 19.07 to 3.41, and the IS score from 64.3 to 288.2, achieving performance levels\\\\n7Preprint\\\\n/uni00000014/uni00000013/uni00000013 /uni00000015/uni00000013/uni00000013 /uni00000016/uni00000013/uni00000013 /uni00000017/uni00000013/uni00000013\\\\n/uni0000002c/uni00000036/uni00000003\\\\n/uni00000018/uni00000014/uni00000013/uni00000014/uni00000018/uni00000029/uni0000002c/uni00000027/uni00000003\\\\n=0\\\\n=104\\\\ns=0\\\\ns=3/uni0000002f/uni0000004f/uni00000044/uni00000050/uni00000044/uni0000002a/uni00000048/uni00000051/uni00000010/uni0000002f\\\\n/uni0000000e/uni00000003/uni00000026/uni00000026/uni00000024\\\\n/uni0000000e/uni00000003/uni00000026/uni00000029/uni0000002a\\\\n/uni00000014/uni00000018/uni00000013 /uni00000015/uni00000013/uni00000013 /uni00000015/uni00000018/uni00000013 /uni00000016/uni00000013/uni00000013 /uni00000016/uni00000018/uni00000013 /uni00000017/uni00000013/uni00000013\\\\n/uni0000002c/uni00000036/uni00000003\\\\n/uni00000015/uni00000016/uni00000017/uni00000018/uni00000019/uni0000001a/uni00000029/uni0000002c/uni00000027/uni00000003\\\\n=0\\\\n=104\\\\ns=0\\\\ns*=3.0s=1.0/uni00000039 /uni00000024/uni00000035/uni00000010/uni00000047/uni00000015/uni00000017\\\\n/uni0000000e/uni00000003/uni00000026/uni00000026/uni00000024\\\\n/uni0000000e/uni00000003/uni00000026/uni00000029/uni0000002a\\\\n/uni0000000e/uni00000003/uni00000026/uni00000029/uni0000002a/uni00000059/uni00000015\\\\nFigure 4: CCA can achieve similar FID-IS trade-offs to CFG by adjusting training parameter λ.\\\\nModel FID ↓ IS sFID ↓Precision Recall\\\\nLlamaGen-L 19.00 64.7 8.78 0.61 0.67\\\\n+DPO 61.69 30.8 44.98 0.36 0.40\\\\n+Unlearn 12.22 111.6 7.99 0.66 0.64\\\\n+CCA 3.43 288.2 7.44 0.81 0.52Model FID↓ IS sFID ↓Precision Recall\\\\nV AR-d24 6.20 154.3 8.50 0.74 0.62\\\\n+DPO 7.53 232.6 19.10 0.85 0.34\\\\n+Unlearn 5.55 165.9 8.41 0.75 0.61\\\\n+CCA 2.63 298.8 7.63 0.84 0.55\\\\nTable 3: Comparision of CCA and LLM alignment algorithms in visual generation.\\\\ncomparable to CFG. This result is compelling, considering that CCA has negligible fine-tunning\\\\noverhead compared with model pretraining and only half of sampling costs compared with CFG.\\\\nFigure 3 presents a qualitative comparison of image samples before and after CCA fine-tuning.\\\\nThe results clearly demonstrate that CCA can vastly improve image fidelity, as well as class-image\\\\nalignment of guidance-free samples.\\\\nTable 2 compares our best-performing models with several state-of-the-art visual generative models.\\\\nWith the help of CCA, we achieve a record-breaking FID score of 2.54 and an IS score of 276.8\\\\nfor guidance-free samples of AR visual models. Although these numbers still somewhat lag behind\\\\nCFG-enhanced performance, they demonstrate the significant potential of alignment algorithms to\\\\nenhance visual generation and indicate the future possibility of replacing guided sampling.\\\\n5.2 C ONTROLLABLE TRADE -OFFS BETWEEN DIVERSITY AND FIDELITY\\\\nA distinctive feature of CFG is its ability to balance diversity and fidelity by adjusting the guidance\\\\nscale. It is reasonable to expect that CCA can achieve a similar trade-off since it essentially targets\\\\nthe same sampling distribution as CFG.\\\\nFigure 4 confirms this expectation: by adjusting the λparameter for fine-tuning, CCA can achieve\\\\nsimilar FID-IS trade-offs to CFG. The key difference is that CCA enhances guidance-free models\\\\nthrough training, while CFG mainly improves the sampling process.\\\\nIt is worth noting that V AR employs a slightly different guidance technique from standard CFG, which\\\\nwe refer to as CFGv2. CFGv2 involves linearly increasing the guidance scale sduring sampling,\\\\nwhich was first proposed by Chang et al. (2023) and found beneficial for certain models. The FID-IS\\\\ncurve of CCA more closely resembles that of standard CFG. Additionally, the hyperparameter βalso\\\\naffects CCA performance. Although our algorithm derivation shows that βis directly related to the\\\\nCFG scale s, we empirically find adjusting βis less effective and less predictable compared with\\\\nadjusting λ. In practice, we typically fix βand adjust λ. We ablate βin Appendix C.\\\\n5.3 C ANLLM A LIGNMENT ALGORITHMS ALSO ENHANCE VISUAL AR?\\\\nIntuitively, existing preference-based LLM alignment algorithms such as DPO and Unlearning\\\\n(Welleck et al., 2019) should also offer improvement for AR visual models given their similarity to\\\\nCCA. However, Table 3 shows that naive applications of these methods fail significantly.\\\\nDPO. As is described in Eq. 5, one can treat negative image-condition pairs as dispreferred data\\\\nand positive ones as preferred data to apply the DPO loss. We ablate βd∈ {0.01,0.1,1.0,10.0}\\\\nand report the best performance in Table 3. Results indicate that DPO fails to enhance pretrained\\\\n8Preprint\\\\n/uni00000013/uni00000011/uni00000013 /uni00000013/uni00000011/uni00000018 /uni00000014/uni00000011/uni00000013 /uni00000015/uni00000011/uni00000013/uni00000016/uni00000011/uni00000013\\\\n/uni0000002a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048/uni00000003/uni00000056/uni00000046/uni00000044/uni0000004f/uni00000048/uni00000003s/uni00000016/uni00000017/uni00000018/uni00000019/uni0000001b/uni00000014/uni00000013/uni00000014/uni00000015/uni00000014/uni00000017/uni00000014/uni00000019 /uni00000029/uni0000002c/uni00000027\\\\n/uni00000029/uni0000002c/uni00000027/uni00000003/uni0000005a/uni00000012/uni00000052/uni00000003/uni00000026/uni00000026/uni00000024\\\\n/uni00000013/uni00000011/uni00000013 /uni00000013/uni00000011/uni00000018 /uni00000014/uni00000011/uni00000013 /uni00000015/uni00000011/uni00000013/uni00000016/uni00000011/uni00000013\\\\n/uni0000002a/uni00000058/uni0000004c/uni00000047/uni00000044/uni00000051/uni00000046/uni00000048/uni00000003/uni00000056/uni00000046/uni00000044/uni0000004f/uni00000048/uni00000003s/uni00000014/uni00000013/uni00000013/uni00000015/uni00000013/uni00000013/uni00000016/uni00000013/uni00000013/uni00000017/uni00000013/uni00000013/uni00000018/uni00000013/uni00000013 /uni0000002c/uni00000036\\\\n/uni0000002c/uni00000036/uni00000003/uni0000005a/uni00000012/uni00000052/uni00000003/uni00000026/uni00000026/uni00000024\\\\n/uni00000014/uni00000011/uni000000133e3/uni00000026/uni00000026/uni00000024/uni00000003/uni00000003/uni00000003/uni00000003\\\\n/uni00000014/uni00000011/uni00000013 /uni00000014/uni00000013 /uni00000014/uni00000013/uni00000013 /uni00000016/uni00000013/uni00000013 /uni00000014/uni00000048/uni00000016 /uni00000016/uni00000048/uni00000016\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni00000003/uni00000003\\\\n/uni00000013/uni00000011/uni00000013/uni00000013/uni00000011/uni00000015/uni00000013/uni00000011/uni00000017/uni00000013/uni00000011/uni00000019 /uni00000032/uni00000053/uni00000057/uni0000004c/uni00000050/uni00000044/uni0000004f/uni00000003/uni00000026/uni00000029/uni0000002a/uni00000003/uni00000056/uni00000046/uni00000044/uni0000004f/uni00000048/uni00000003s*/uni00000003/uni0000000b/uni00000049/uni00000052/uni00000055/uni00000003/uni00000029/uni0000002c/uni00000027/uni0000000c /uni0000005a/uni00000012/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000048/uni00000047/uni00000003/uni00000056/uni00000044/uni00000050/uni00000053/uni0000004f/uni0000004c/uni00000051/uni0000004a\\\\n/uni0000005a/uni00000012/uni00000052/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000048/uni00000047/uni00000003/uni00000056/uni00000044/uni00000050/uni00000053/uni0000004f/uni0000004c/uni00000051/uni0000004a\\\\n/uni00000014/uni00000011/uni00000013 /uni00000014/uni00000013 /uni00000014/uni00000013/uni00000013 /uni00000016/uni00000013/uni00000013 /uni00000014/uni00000048/uni00000016 /uni00000016/uni00000048/uni00000016\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni00000003/uni00000003\\\\n/uni00000015/uni00000011/uni00000018/uni00000016/uni00000017/uni00000019/uni0000001b/uni00000014/uni00000015 /uni00000032/uni00000053/uni00000057/uni0000004c/uni00000050/uni00000044/uni0000004f/uni00000003/uni00000029/uni0000002c/uni00000027\\\\n*/uni00000003/uni00000049/uni00000052/uni00000055/uni00000003/uni00000026/uni00000026/uni00000024/uni00000003/uni00000052/uni00000051/uni0000004f/uni0000005c\\\\n*/uni00000003/uni00000049/uni00000052/uni00000055/uni00000003/uni00000026/uni00000026/uni00000024/uni0000000e/uni00000026/uni00000029/uni0000002a\\\\n/uni00000029/uni0000002c/uni00000027/uni00000003/uni0000005a/uni00000012/uni00000052/uni00000003/uni00000026/uni00000026/uni00000024/uni0000005a/uni00000012/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000048/uni00000047/uni00000003/uni00000056/uni00000044/uni00000050/uni00000053/uni0000004f/uni0000004c/uni00000051/uni0000004a\\\\n/uni0000005a/uni00000012/uni00000052/uni00000003/uni0000004a/uni00000058/uni0000004c/uni00000047/uni00000048/uni00000047/uni00000003/uni00000056/uni00000044/uni00000050/uni00000053/uni0000004f/uni0000004c/uni00000051/uni0000004a\\\\nFigure 5: The impact of training parameter λon the performance of CCA+CFG.\\\\n/uni00000014/uni00000014/uni00000014/uni00000030 /uni00000016/uni00000017/uni00000016/uni00000030 /uni0000001a/uni0000001a/uni00000018/uni00000030 /uni00000014/uni00000011/uni00000017/uni00000025 /uni00000016/uni00000011/uni00000014/uni00000025/uni00000015/uni00000016/uni00000017/uni00000018/uni00000019/uni0000001a/uni00000029/uni0000002c/uni00000027/uni00000003\\\\n/uni0000001a/uni00000011/uni00000013/uni00000017\\\\n/uni00000016/uni00000011/uni00000017/uni00000016/uni00000016/uni00000011/uni00000014 /uni00000016/uni00000011/uni00000014/uni00000015\\\\n/uni00000015/uni00000011/uni00000019/uni0000001c/uni00000019/uni00000011/uni00000013/uni0000001c\\\\n/uni00000016/uni00000011/uni00000013/uni0000001b\\\\n/uni00000015/uni00000011/uni00000019/uni00000015\\\\n/uni00000015/uni00000011/uni00000016/uni00000017\\\\n/uni00000015/uni00000011/uni00000014/uni0000001b/uni00000018/uni00000011/uni00000015/uni00000018\\\\n/uni00000015/uni00000011/uni0000001a/uni00000015\\\\n/uni00000015/uni00000011/uni00000016/uni00000013\\\\n/uni00000015/uni00000011/uni00000013/uni00000019/uni00000015/uni00000011/uni00000015/uni0000001b/uni00000026/uni00000026/uni00000024/uni0000000e/uni00000026/uni00000029/uni0000002a\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni00000052/uni00000051/uni0000004f/uni0000005c\\\\n/uni00000026/uni00000029/uni0000002a/uni00000003/uni00000052/uni00000051/uni0000004f/uni0000005c\\\\n/uni00000014/uni00000014/uni00000014/uni00000030 /uni00000016/uni00000017/uni00000016/uni00000030 /uni0000001a/uni0000001a/uni00000018/uni00000030 /uni00000014/uni00000011/uni00000017/uni00000025 /uni00000016/uni00000011/uni00000014/uni00000025/uni00000014/uni0000001a/uni00000018/uni00000015/uni00000013/uni00000013/uni00000015/uni00000015/uni00000018/uni00000015/uni00000018/uni00000013/uni00000015/uni0000001a/uni00000018/uni00000016/uni00000013/uni00000013/uni00000016/uni00000015/uni00000018/uni0000002c/uni00000036/uni00000003\\\\n/uni00000015/uni00000015/uni00000013/uni00000011/uni0000001a/uni00000015/uni0000001b/uni0000001b/uni00000011/uni00000015/uni00000016/uni00000016/uni0000001c/uni00000011/uni00000013\\\\n/uni00000016/uni00000015/uni0000001c/uni00000011/uni0000001b\\\\n/uni00000015/uni0000001a/uni00000019/uni00000011/uni0000001b\\\\n/uni00000014/uni0000001b/uni00000015/uni00000011/uni00000018/uni00000015/uni00000018/uni00000019/uni00000011/uni00000014 /uni00000015/uni00000017/uni00000017/uni00000011/uni00000014/uni00000015/uni00000018/uni00000016/uni00000011/uni0000001c/uni00000015/uni00000019/uni00000016/uni00000011/uni00000016\\\\n/uni00000015/uni00000013/uni00000017/uni00000011/uni0000001a/uni00000015/uni00000019/uni00000017/uni00000011/uni00000017/uni00000015/uni0000001a/uni00000013/uni00000011/uni00000016/uni00000015/uni0000001b/uni00000017/uni00000011/uni0000001a /uni00000015/uni0000001b/uni00000015/uni00000011/uni00000016\\\\n/uni00000026/uni00000026/uni00000024/uni0000000e/uni00000026/uni00000029/uni0000002a\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni00000052/uni00000051/uni0000004f/uni0000005c\\\\n/uni00000026/uni00000029/uni0000002a/uni00000003/uni00000052/uni00000051/uni0000004f/uni0000005c\\\\nFigure 6: Integration of CCA+CFG yields improved performance over CFG alone.\\\\nmodels, even causing performance collapse for LlamaGen-L. By inspecting training logs, we find\\\\nthat the likelihood of the positive data continuously decreases during fine-tuning, which may explain\\\\nthe collapse. This phenomenon is a well-observed problem of DPO (Chen et al., 2024a; Pal et al.,\\\\n2024), stemming from its focus on optimizing only the relative likelihood between preferred and\\\\ndispreferred data, rather than controlling likelihood for positive and negative image-condition pairs\\\\nseparately. We refer interested readers to Chen et al. (2024a) for a detailed discussion.\\\\nUnlearning. Also known as unlikelihood training, this method maximizes logpθ(x|c)through\\\\nstandard maximum likelihood training on positive data, while minimizing logpθ(x|cneg)tounlearn\\\\nnegative data. A training parameter λucontrols the weight of the unlearning loss. We find that with\\\\nsmall 0.01≤λu≤0.1, Unlearning provides some benefit, but it is far less effective than CCA. This\\\\nsuggests the necessity of including a frozen reference model.\\\\n5.4 I NTEGRATION OF CCA AND CFG\\\\nIf extra sampling costs and design inconsistencies of CFG are not concerns, could CCA still be\\\\nhelpful? A takeaway conclusion is yes: CCA+CFG generally outperforms CFG (Figure 6), but it\\\\nrequires distinct hyperparameter choices compared with CCA-only training.\\\\nImplementation. After pretraining the unconditional AR visual model by randomly dropping\\\\nconditions, CFG requires us to also fine-tune the unconditional model during alignment. To achieve\\\\nthis, we follow previous approaches by also randomly replacing data conditions with [MASK] tokens\\\\nat a probability of 10%. These unconditional samples are treated as positive data during CCA training.\\\\nComparison of CCA-only and CCA+CFG. They require different hyperparameters. As shown\\\\nin Figure 5, a larger λis typically needed for optimal FID scores in guidance-free generation. For\\\\nmodels optimized for guidance-free sampling, adding CFG guidance does not further reduce the FID\\\\nscore. However, with a smaller λ, CCA+CFG could outperform the CFG method.\\\\n6 R ELATED WORKS\\\\nVisual generative models. Generative adversarial networks (GANs) (Goodfellow et al., 2014;\\\\nBrock et al., 2018; Karras et al., 2019; Kang et al., 2023) and diffusion models (Ho et al., 2020;\\\\n9Preprint\\\\nSong & Ermon, 2019; Song et al., 2020; Dhariwal & Nichol, 2021; Kingma & Gao, 2024) are\\\\nrepresentative modeling methods for visual content generation, widely recognized for their ability to\\\\nproduce realistic and artistic images (Sauer et al., 2022; Ho et al., 2022; Ramesh et al., 2022; Podell\\\\net al., 2023). However, because these methods are designed for continuous data like images, they\\\\nstruggle to effectively model discrete data such as text, limiting the development of unified multimodal\\\\nmodels for both vision and language. Recent approaches seek to address this by integrating diffusion\\\\nmodels with language models (Team, 2024; Li et al., 2024; Zhou et al., 2024). Another line of works\\\\n(Chang et al., 2022; 2023; Yu et al., 2023; Xie et al., 2024; Ramesh et al., 2021; Yu et al., 2022)\\\\nexplores discretizing images (Van Den Oord et al., 2017; Esser et al., 2021) and directly applying\\\\nlanguage models such as BERT-style (Devlin et al., 2018) masked-prediction models and GPT-style\\\\n(Radford et al., 2018) autoregressive models for image generation.\\\\nLanguage model alignment. Different from visual generative models which generally enhance\\\\nsample quality through sampling-based methods (Dhariwal & Nichol, 2021; Ho & Salimans, 2022;\\\\nZhao et al., 2022; Lu et al., 2023), LLMs primarily employ training-based alignment techniques\\\\nto improve instruction-following abilities (Touvron et al., 2023b; OpenAI, 2023). Reinforcement\\\\nLearning (RL) is well-suited for aligning LLMs with human feedback (Schulman et al., 2017; Ouyang\\\\net al., 2022). However, this method requires learning a reward model before optimizing LLMs, leading\\\\nto an indirect two-stage optimization process. Recent developments in alignment techniques (Rafailov\\\\net al., 2023; Cai et al., 2023; Azar et al., 2024; Ethayarajh et al., 2024; Chen et al., 2024a; Ji et al.,\\\\n2024) have streamlined this process. They enable direct alignment of LMs through a singular\\\\nloss. Among all LLM alignment algorithms, our method is perhaps most similar to NCA (Chen\\\\net al., 2024a). Both NCA and CCA are theoretically grounded in the NCE framework (Gutmann &\\\\nHyvärinen, 2012). Their differences are mainly empirical regarding loss implementations, particularly\\\\nin how to estimate expectations under the product of two marginal distributions.\\\\nVisual alignment. Motivated by the success of alignment techniques in LLMs, several studies\\\\nhave also investigated aligning visual generative models with human preferences using RL (Black\\\\net al., 2023a; Xu et al., 2024) or DPO (Black et al., 2023b; Wallace et al., 2023). For diffusion\\\\nmodels, the application is not straightforward and must rely on some theoretical approximations, as\\\\ndiffusion models do not allow direct likelihood calculation, which is required by most LLM alignment\\\\nalgorithms (Chen et al., 2024b). Moreover, previous attempts at visual alignment have primarily\\\\nfocused on enhancing the aesthetic quality of generated images and necessitate a different dataset\\\\nfrom the pretrained one. Our work distinguishes itself from prior research by having a fundamentally\\\\ndifferent optimization objective (replacing CFG) and does not rely on any additional data input.\\\\n7 C ONCLUSION\\\\nIn this paper, we propose Condition Contrastive Alignment (CCA) as a fine-tuning algorithm for\\\\nAR visual generation models. CCA can significantly enhance the guidance-free sample quality\\\\nof pretrained models without any modification of the sampling process. This paves the way for\\\\nfurther development in multimodal generative models and cuts the cost of AR visual generation by\\\\nhalf in comparison to CFG. Our research also highlights the strong theoretical connection between\\\\nlanguage-targeted alignment and visual-targeted guidance methods, facilitating future research of\\\\nunifying visual modeling and language modeling.\\\\nACKNOWLEDGMENTS\\\\nWe thank Fan Bao, Kai Jiang, Xiang Li, and Min Zhao for providing valuable suggestions. We thank\\\\nKeyu Tian and Kaiwen Zheng for the discussion.\\\\n10Preprint\\\\nREFERENCES\\\\nMohammad Gheshlaghi Azar, Zhaohan Daniel Guo, Bilal Piot, Remi Munos, Mark Rowland, Michal\\\\nValko, and Daniele Calandriello. A general theoretical paradigm to understand learning from\\\\nhuman preferences. In International Conference on Artificial Intelligence and Statistics , pp.\\\\n4447–4455. PMLR, 2024.\\\\nFan Bao, Shen Nie, Kaiwen Xue, Yue Cao, Chongxuan Li, Hang Su, and Jun Zhu. All are worth\\\\nwords: A vit backbone for diffusion models. In Proceedings of the IEEE/CVF conference on\\\\ncomputer vision and pattern recognition , pp. 22669–22679, 2023.\\\\nKevin Black, Michael Janner, Yilun Du, Ilya Kostrikov, and Sergey Levine. Training diffusion models\\\\nwith reinforcement learning. arXiv preprint arXiv:2305.13301 , 2023a.\\\\nKevin Black, Michael Janner, Yilun Du, Ilya Kostrikov, and Sergey Levine. Training diffusion models\\\\nwith reinforcement learning. arXiv preprint arXiv:2305.13301 , 2023b.\\\\nRalph Allan Bradley and Milton E Terry. Rank analysis of incomplete block designs: I. the method\\\\nof paired comparisons. Biometrika , 39(3/4):324–345, 1952.\\\\nAndrew Brock, Jeff Donahue, and Karen Simonyan. Large scale gan training for high fidelity natural\\\\nimage synthesis. arXiv preprint arXiv:1809.11096 , 2018.\\\\nTianchi Cai, Xierui Song, Jiyan Jiang, Fei Teng, Jinjie Gu, and Guannan Zhang. Ulma: Unified\\\\nlanguage model alignment with demonstration and point-wise human preference. arXiv preprint\\\\narXiv:2312.02554 , 2023.\\\\nHuiwen Chang, Han Zhang, Lu Jiang, Ce Liu, and William T Freeman. Maskgit: Masked generative\\\\nimage transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern\\\\nRecognition , pp. 11315–11325, 2022.\\\\nHuiwen Chang, Han Zhang, Jarred Barber, AJ Maschinot, Jose Lezama, Lu Jiang, Ming-Hsuan Yang,\\\\nKevin Murphy, William T Freeman, Michael Rubinstein, et al. Muse: Text-to-image generation\\\\nvia masked generative transformers. arXiv preprint arXiv:2301.00704 , 2023.\\\\nHuayu Chen, Guande He, Hang Su, and Jun Zhu. Noise contrastive alignment of language models\\\\nwith explicit rewards. Advances in neural information processing systems , 2024a.\\\\nHuayu Chen, Kaiwen Zheng, Hang Su, and Jun Zhu. Aligning diffusion behaviors with q-functions\\\\nfor efficient continuous control. Advances in neural information processing systems , 2024b.\\\\nJacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep\\\\nbidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805 , 2018.\\\\nPrafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances\\\\nin neural information processing systems , 34:8780–8794, 2021.\\\\nPatrick Esser, Robin Rombach, and Bjorn Ommer. Taming transformers for high-resolution image\\\\nsynthesis. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition ,\\\\npp. 12873–12883, 2021.\\\\nKawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, and Douwe Kiela. Kto: Model\\\\nalignment as prospect theoretic optimization. arXiv preprint arXiv:2402.01306 , 2024.\\\\nShanghua Gao, Pan Zhou, Ming-Ming Cheng, and Shuicheng Yan. Masked diffusion transformer is a\\\\nstrong image synthesizer. In Proceedings of the IEEE/CVF International Conference on Computer\\\\nVision , pp. 23164–23173, 2023.\\\\nIan Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair,\\\\nAaron Courville, and Yoshua Bengio. Generative adversarial nets. Advances in neural information\\\\nprocessing systems , 27, 2014.\\\\nMichael U Gutmann and Aapo Hyvärinen. Noise-contrastive estimation of unnormalized statistical\\\\nmodels, with applications to natural image statistics. Journal of machine learning research , 13(2),\\\\n2012.\\\\n11Preprint\\\\nMartin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans\\\\ntrained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural\\\\ninformation processing systems , 30, 2017.\\\\nJonathan Ho and Tim Salimans. Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598 ,\\\\n2022.\\\\nJonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in\\\\nneural information processing systems , 33:6840–6851, 2020.\\\\nJonathan Ho, Chitwan Saharia, William Chan, David J Fleet, Mohammad Norouzi, and Tim Salimans.\\\\nCascaded diffusion models for high fidelity image generation. The Journal of Machine Learning\\\\nResearch , 23(1):2249–2281, 2022.\\\\nHaozhe Ji, Cheng Lu, Yilin Niu, Pei Ke, Hongning Wang, Jun Zhu, Jie Tang, and Minlie\\\\nHuang. Towards efficient and exact optimization of language model alignment. arXiv preprint\\\\narXiv:2402.00856 , 2024.\\\\nMinguk Kang, Jun-Yan Zhu, Richard Zhang, Jaesik Park, Eli Shechtman, Sylvain Paris, and Taesung\\\\nPark. Scaling up gans for text-to-image synthesis. In Proceedings of the IEEE/CVF Conference on\\\\nComputer Vision and Pattern Recognition , pp. 10124–10134, 2023.\\\\nTero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative\\\\nadversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern\\\\nrecognition , pp. 4401–4410, 2019.\\\\nDiederik Kingma and Ruiqi Gao. Understanding diffusion objectives as the elbo with simple data\\\\naugmentation. Advances in Neural Information Processing Systems , 36, 2024.\\\\nDoyup Lee, Chiheon Kim, Saehoon Kim, Minsu Cho, and Wook-Shin Han. Autoregressive image\\\\ngeneration using residual quantization. In Proceedings of the IEEE/CVF Conference on Computer\\\\nVision and Pattern Recognition , pp. 11523–11532, 2022.\\\\nTianhong Li, Huiwen Chang, Shlok Mishra, Han Zhang, Dina Katabi, and Dilip Krishnan. Mage:\\\\nMasked generative encoder to unify representation learning and image synthesis. In Proceedings\\\\nof the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. 2142–2152, 2023.\\\\nTianhong Li, Yonglong Tian, He Li, Mingyang Deng, and Kaiming He. Autoregressive image\\\\ngeneration without vector quantization. arXiv preprint arXiv:2406.11838 , 2024.\\\\nCheng Lu, Huayu Chen, Jianfei Chen, Hang Su, Chongxuan Li, and Jun Zhu. Contrastive energy\\\\nprediction for exact energy-guided diffusion sampling in offline reinforcement learning. In\\\\nInternational Conference on Machine Learning , pp. 22825–22855. PMLR, 2023.\\\\nOpenAI. Gpt-4 technical report. arXiv preprint arXiv:2303.08774 , 2023.\\\\nLong Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong\\\\nZhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow\\\\ninstructions with human feedback. Advances in neural information processing systems , 35:27730–\\\\n27744, 2022.\\\\nArka Pal, Deep Karkhanis, Samuel Dooley, Manley Roberts, Siddartha Naidu, and Colin White.\\\\nSmaug: Fixing failure modes of preference optimisation with dpo-positive. arXiv preprint\\\\narXiv:2402.13228 , 2024.\\\\nWilliam Peebles and Saining Xie. Scalable diffusion models with transformers. In Proceedings of\\\\nthe IEEE/CVF International Conference on Computer Vision , pp. 4195–4205, 2023.\\\\nDustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas Müller, Joe\\\\nPenna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image\\\\nsynthesis. arXiv preprint arXiv:2307.01952 , 2023.\\\\nAlec Radford, Karthik Narasimhan, Tim Salimans, Ilya Sutskever, et al. Improving language\\\\nunderstanding by generative pre-training. article , 2018.\\\\n12Preprint\\\\nRafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea\\\\nFinn. Direct preference optimization: Your language model is secretly a reward model. In\\\\nThirty-seventh Conference on Neural Information Processing Systems , 2023.\\\\nAditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen,\\\\nand Ilya Sutskever. Zero-shot text-to-image generation. In International Conference on Machine\\\\nLearning , pp. 8821–8831. PMLR, 2021.\\\\nAditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-\\\\nconditional image generation with clip latents. arXiv preprint arXiv:2204.06125 , 1(2):3, 2022.\\\\nRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-\\\\nresolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF confer-\\\\nence on computer vision and pattern recognition , pp. 10684–10695, 2022.\\\\nTim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, and Xi Chen.\\\\nImproved techniques for training gans. Advances in neural information processing systems , 29,\\\\n2016.\\\\nAxel Sauer, Katja Schwarz, and Andreas Geiger. Stylegan-xl: Scaling stylegan to large diverse\\\\ndatasets. In ACM SIGGRAPH 2022 conference proceedings , pp. 1–10, 2022.\\\\nJohn Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy\\\\noptimization algorithms. arXiv preprint arXiv:1707.06347 , 2017.\\\\nJohn Schulman, Barret Zoph, Christina Kim, Jacob Hilton, Jacob Menick, Jiayi Weng, Juan Fe-\\\\nlipe Ceron Uribe, Liam Fedus, Luke Metz, Michael Pokorny, et al. Chatgpt: Optimizing language\\\\nmodels for dialogue. OpenAI blog , 2022.\\\\nJiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv\\\\npreprint arXiv:2010.02502 , 2020.\\\\nYang Song and Stefano Ermon. Generative modeling by estimating gradients of the data distribution.\\\\nAdvances in neural information processing systems , 32, 2019.\\\\nPeize Sun, Yi Jiang, Shoufa Chen, Shilong Zhang, Bingyue Peng, Ping Luo, and Zehuan Yuan.\\\\nAutoregressive model beats diffusion: Llama for scalable image generation. arXiv preprint\\\\narXiv:2406.06525 , 2024.\\\\nChameleon Team. Chameleon: Mixed-modal early-fusion foundation models. arXiv preprint\\\\narXiv:2405.09818 , 2024.\\\\nKeyu Tian, Yi Jiang, Zehuan Yuan, Bingyue Peng, and Liwei Wang. Visual autoregressive modeling:\\\\nScalable image generation via next-scale prediction. arXiv preprint arXiv:2404.02905 , 2024.\\\\nHugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée\\\\nLacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and\\\\nefficient foundation language models. arXiv preprint arXiv:2302.13971 , 2023a.\\\\nHugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay\\\\nBashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation\\\\nand fine-tuned chat models. arXiv preprint arXiv:2307.09288 , 2023b.\\\\nAaron Van Den Oord, Oriol Vinyals, et al. Neural discrete representation learning. Advances in\\\\nneural information processing systems , 30, 2017.\\\\nBram Wallace, Meihua Dang, Rafael Rafailov, Linqi Zhou, Aaron Lou, Senthil Purushwalkam,\\\\nStefano Ermon, Caiming Xiong, Shafiq Joty, and Nikhil Naik. Diffusion model alignment using\\\\ndirect preference optimization. arXiv preprint arXiv:2311.12908 , 2023.\\\\nSean Welleck, Ilia Kulikov, Stephen Roller, Emily Dinan, Kyunghyun Cho, and Jason Weston. Neural\\\\ntext generation with unlikelihood training. arXiv preprint arXiv:1908.04319 , 2019.\\\\n13Preprint\\\\nJinheng Xie, Weijia Mao, Zechen Bai, David Junhao Zhang, Weihao Wang, Kevin Qinghong Lin,\\\\nYuchao Gu, Zhijie Chen, Zhenheng Yang, and Mike Zheng Shou. Show-o: One single transformer\\\\nto unify multimodal understanding and generation. arXiv preprint arXiv:2408.12528 , 2024.\\\\nJiazheng Xu, Xiao Liu, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, and Yuxiao Dong.\\\\nImagereward: Learning and evaluating human preferences for text-to-image generation. Advances\\\\nin Neural Information Processing Systems , 36, 2024.\\\\nJiahui Yu, Xin Li, Jing Yu Koh, Han Zhang, Ruoming Pang, James Qin, Alexander Ku, Yuanzhong\\\\nXu, Jason Baldridge, and Yonghui Wu. Vector-quantized image modeling with improved vqgan.\\\\narXiv preprint arXiv:2110.04627 , 2021.\\\\nJiahui Yu, Yuanzhong Xu, Jing Yu Koh, Thang Luong, Gunjan Baid, Zirui Wang, Vijay Vasudevan,\\\\nAlexander Ku, Yinfei Yang, Burcu Karagol Ayan, et al. Scaling autoregressive models for content-\\\\nrich text-to-image generation. arXiv preprint arXiv:2206.10789 , 2(3):5, 2022.\\\\nLijun Yu, José Lezama, Nitesh B Gundavarapu, Luca Versari, Kihyuk Sohn, David Minnen, Yong\\\\nCheng, Agrim Gupta, Xiuye Gu, Alexander G Hauptmann, et al. Language model beats diffusion–\\\\ntokenizer is key to visual generation. arXiv preprint arXiv:2310.05737 , 2023.\\\\nMin Zhao, Fan Bao, Chongxuan Li, and Jun Zhu. Egsde: Unpaired image-to-image translation\\\\nvia energy-guided stochastic differential equations. Advances in Neural Information Processing\\\\nSystems , 35:3609–3623, 2022.\\\\nChunting Zhou, Lili Yu, Arun Babu, Kushal Tirumala, Michihiro Yasunaga, Leonid Shamis, Jacob\\\\nKahn, Xuezhe Ma, Luke Zettlemoyer, and Omer Levy. Transfusion: Predict the next token and\\\\ndiffuse images with one multi-modal model. arXiv preprint arXiv:2408.11039 , 2024.\\\\n14Preprint\\\\nw/o Guidance +CCA (w/o Guidance) w/ CFG Guidance\\\\nFigure 7: Comparison of LlamaGen-L samples generated with CCA or CFG.\\\\n15Preprint\\\\nw/o Guidance +CCA (w/o Guidance) w/ CFG Guidance\\\\nFigure 8: Comparison of V AR-d24 samples generated with CCA or CFG.\\\\n16Preprint\\\\nA T HEORETICAL PROOFS\\\\nIn this section, we provide the proof of Theorem 3.1.\\\\nTheorem A.1 (Noise Contrastive Estimation ) .Letrθbe a parameterized model which takes in an\\\\nimage-condition pair (x,c)and outputs a scalar value rθ(x,c). Consider the loss function:\\\\nLNCE\\\\nθ(x,c) =−Ep(x,c)logσ(rθ(x,c))−Ep(x)p(c)logσ(−rθ(x,c)). (13)\\\\nGiven unlimited model expressivity for rθ, the optimal solution for minimizing LNCE\\\\nθsatisfies\\\\nr∗\\\\nθ(x,c) = logp(x|c)\\\\np(x). (14)\\\\nProof. First, we construct two binary (Bernoulli) distributions:\\\\nQx,c:={p(x,c)\\\\np(x,c) +p(x)p(c),p(x)p(c)\\\\np(x,c) +p(x)p(c)}={p(x|c)\\\\np(x|c) +p(x),p(x)\\\\np(x|c) +p(x)}\\\\nPθ\\\\nx,c:={erθ(x,c)\\\\nerθ(x,c)+ 1,1\\\\nerθ(x,c)+ 1}={σ(rθ(x,c)),1−σ(rθ(x,c))}\\\\nThen we rewrite LNCE\\\\nθ(x,c)as\\\\nLNCE\\\\nθ(x,c) =−Ep(x,c)logσ(rθ(x,c))−Ep(x)p(c)logσ(−rθ(x,c))\\\\n=−Zh\\\\np(x,c) logσ(rθ(x,c)) +p(x)p(c) logσ(−rθ(x,c))i\\\\ndxdc\\\\n=−Zh\\\\n(p(x,c) +p(x)p(c))i\\\\nhp(x,c)\\\\np(x,c) +p(x)p(c)logσ(rθ(x,c)) +p(x)p(c)\\\\np(x,c) +p(x)p(c)log\\\\x02\\\\n1−σ(rθ(x,c))\\\\x03i\\\\ndxdc\\\\n=Zh\\\\n(p(x,c) +p(x)p(c))i\\\\nH(Qx,c, Pθ\\\\nx,c)dxdc\\\\n=Zh\\\\n(p(x,c) +p(x)p(c))ih\\\\nDKL(Qx,c∥Pθ\\\\nx,c) +H(Qx,c)i\\\\ndxdc\\\\nHere H(Qx,c, Pθ\\\\nx,c)represents the cross-entropy between distributions Qx,candPθ\\\\nx,c.H(Qx,c)is\\\\nthe entropy of Qx,c, which can be regarded as a constant number with respect to parameter θ. Due to\\\\nthe theoretical properties of KL-divergence, we have\\\\nLNCE\\\\nθ(x,c) =Zh\\\\n(p(x,c) +p(x)p(c))ih\\\\nDKL(Qx,c∥Pθ\\\\nx,c) +H(Qx,c)i\\\\ndxdc\\\\n≥Zh\\\\n(p(x,c) +p(x)p(c))i\\\\nH(Qx,c)dxdc\\\\nconstantly hold. The equality holds if and only if Qx,c=Pθ\\\\nx,c, such that\\\\nσ(rθ(x,c)) =erθ(x,c)\\\\nerθ(x,c)+ 1=p(x,c)\\\\np(x,c) +p(x)p(c)\\\\nrθ(x,c) = logp(x,c)\\\\np(x)p(c)= logp(x|c)\\\\np(x)\\\\n17Preprint\\\\nB T HEORETICAL ANALYSES OF THE NORMALIZING CONSTANT\\\\nWe omit a normalizing constant in Eq. 7 for brevity when deriving CCA. Strictly speaking, the target\\\\nsampling distribution should be:\\\\npsample(x|c) =1\\\\nZ(c)p(x|c)[p(x|c)\\\\np(x)]s,\\\\nsuch that\\\\n1\\\\nslogpsample(x|c)\\\\np(x|c)= logp(x|c)\\\\np(x)−1\\\\nslogZ(c).\\\\nThe normalizing constant Z(c)ensures that psample(x|c)is properly normalized, i.e.,R\\\\npsample(x|c)dx= 1. We have Z(c) =R\\\\np(x|c)[p(x|c)\\\\np(x)]sdx=Ep(x|c)[p(x|c)\\\\np(x)]s.\\\\nTo mitigate the additional effects introduced by Z(c), in our practical algorithm, we introduce a new\\\\ntraining parameter λto bias the optimal solution for Noise Contrastive Estimation. Below, we present\\\\na result that is stronger than Theorem 3.1.\\\\nTheorem B.1. Letλc>0be a scalar function conditioned only on c. Consider the loss function:\\\\nLNCE\\\\nθ(x,c) =−Ep(x,c)logσ(rθ(x,c))−λcEp(x)p(c)logσ(−rθ(x,c)). (15)\\\\nGiven unlimited model expressivity for rθ, the optimal solution for minimizing LNCE\\\\nθsatisfies\\\\nr∗\\\\nθ(x,c) = logp(x|c)\\\\np(x)−logλc. (16)\\\\nProof. We omit the full proof here, as it requires only a redefinition of the distributions Qx,cfrom\\\\nthe proof of Theorem A.1:\\\\nQx,c:={p(x,c)\\\\np(x,c) +λcp(x)p(c),λcp(x)p(c)\\\\np(x,c) +λcp(x)p(c)}={p(x|c)\\\\np(x|c) +λcp(x),λcp(x)\\\\np(x|c) +λcp(x)}\\\\nThen we can follow the steps in the proof of Theorem A.1 to arrive at the result.\\\\nIf let λc:=Z(c)1\\\\ns=\\\\x02\\\\nEp(x|c)[p(x|c)\\\\np(x)]s\\\\x031\\\\ns, we could guarantee the convergence of psample\\\\nθ topsample.\\\\nHowever, in practice estimating Z(c)could be intricately difficult, so we formalize λcas a training\\\\nparameter, resulting in our practical algorithm in Eq. 12.\\\\nC A DDITIONAL EXPERIMENTAL RESULTS\\\\nWe provide more image samples to compare CCA and CFG in Figure 7 and Figure 8.\\\\nWe illustrate the effect of training parameter βon the FID-IS trade-off in Figure 9. Overall, βaffects\\\\nthe fidelity-diversity trade-off similar to CCA λand the CFG method.\\\\n/uni00000018/uni00000013 /uni00000014/uni00000013/uni00000013 /uni00000014/uni00000018/uni00000013 /uni00000015/uni00000013/uni00000013 /uni00000015/uni00000018/uni00000013 /uni00000016/uni00000013/uni00000013 /uni00000016/uni00000018/uni00000013 /uni00000017/uni00000013/uni00000013\\\\n/uni0000002c/uni00000036/uni00000018/uni00000014/uni00000013/uni00000014/uni00000018/uni00000029/uni0000002c/uni00000027\\\\n/uni00000030/uni00000052/uni00000047/uni00000048/uni0000004f\\\\n/uni00000026/uni00000026/uni00000024/uni00000003/uni0000000b/uni0000004"}}},{"rowIdx":75,"cells":{"Researcher Name":{"kind":"string","value":"Justin C. Ondry"},"ORCID":{"kind":"string","value":"0000-0001-9113-3420"},"Researcher Query Keyword":{"kind":"string","value":"High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals"},"Research Full Paper":{"kind":"string","value":"{}"},"Similar Full Paper":{"kind":"string","value":"{}"}}},{"rowIdx":76,"cells":{"Researcher Name":{"kind":"string","value":"Kailai Lin"},"ORCID":{"kind":"string","value":"0000-0001-8509-5560"},"Researcher Query Keyword":{"kind":"string","value":"High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals"},"Research Full Paper":{"kind":"string","value":"{}"},"Similar Full Paper":{"kind":"string","value":"{}"}}},{"rowIdx":77,"cells":{"Researcher Name":{"kind":"string","value":"Benjamin F. Hammel"},"ORCID":{"kind":"string","value":"0000-0002-9155-9875"},"Researcher Query Keyword":{"kind":"string","value":"High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals"},"Research Full Paper":{"kind":"string","value":"{}"},"Similar Full Paper":{"kind":"string","value":"{}"}}},{"rowIdx":78,"cells":{"Researcher Name":{"kind":"string","value":"Di Wang"},"ORCID":{"kind":"string","value":"0000-0002-5841-5060"},"Researcher Query Keyword":{"kind":"string","value":"High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals"},"Research Full Paper":{"kind":"string","value":"{}"},"Similar Full Paper":{"kind":"string","value":"{}"}}},{"rowIdx":79,"cells":{"Researcher Name":{"kind":"string","value":"Stella Chariton"},"ORCID":{"kind":"string","value":"0000-0001-5522-0498"},"Researcher Query Keyword":{"kind":"string","value":"High-Pressure Yttrium Hydrides Synthesis and Characterization"},"Research Full Paper":{"kind":"string","value":"{'Superconductivity up to 243\\\\u2009K in the yttrium-hydrogen system under high pressure': 'Title: Superconductivity up to 243\\\\u2009K in the yttrium-hydrogen system under high pressure\\\\nSuperconductivity up to 243 K in yttrium hydrides under high pressure \\\\n \\\\nP. P. Kong1, V. S. Minkov1, M. A. Kuzovnikov1,5, S. P. Besedin1, A. P. Drozdov1, S. Mozaffari2, L. \\\\nBalicas2, F.F. Balakirev3, V. B. Prakapenka4, E. Greenberg4, D. A. Knyazev1 and M. I. Eremets1* \\\\n \\\\n1Max-Planck Institut für Chemie, Hahn-Meitner Weg 1, 55128 Mainz, Germany \\\\n2National High Magnetic Field Laboratory, Florida State University, Tallahassee, Florida 32310, USA \\\\n3NHMFL, Los Alamos National Laboratory, MS E536, Los Alamos, New Mexico 87545, USA \\\\n4Center for Advanced Radiation Sources, University of Chicago, 5640 South Ellis Avenue, Chicago, \\\\nIllinois, 60637, USA \\\\n5Institute of Solid State Physics Russian Academy of Sciences, 2 Academician Ossipyan str., \\\\nChernogolovka, Moscow District 142432, Russia \\\\n \\\\n \\\\nThe discovery of high-temperature conventional superconductivity in H3S with a critical temperature of \\\\nTc=203 K1 was followed by the recent record of Tc ~250 K in the face-centered cubic (fcc) lanthanum \\\\nhydride LaH102,3 compound. It was realized in a new class of hydrogen-dominated compounds having a \\\\nclathrate-like crystal structure in which hydrogen atoms form a 3D framework and surround a host atom of \\\\nrare earth elements4,5. Yttrium hydrides are predicted to have even higher Tc exceeding room temperature. \\\\nIn this paper, we synthesized and refined the crystal structure of new hydrides: YH4, YH6, and YH9 at \\\\npressures up to 237 GPa finding that YH4 crystalizes in the I4/mmm lattice, YH6 in Im-3m lattice and YH9 \\\\nin P63/mmc lattice in excellent agreement with the calculations5,6. The observed very high-temperature \\\\nsuperconductivity is comparable to that found in fcc-LaH102: the pressure dependence of Tc for YH9 also \\\\ndisplays a “dome like shape” with the highest Tc of 243 K at 201 GPa. We also observed a Tc of 227 K at \\\\n237 GPa for the YH6 phase. However, the measured Tcs are notably lower by \\\\uf07e30 K than predicted5,6. \\\\nEvidence for superconductivity includes the observation of zero electrical resistance, a decrease of Tc under \\\\nan external magnetic field and an isotope effect. The theoretically predicted fcc YH10 with the promising \\\\nhighest Tc>300 K was not stabilized in our experiments under pressures up to 237 GPa. \\\\n \\\\n \\\\n \\\\n \\\\n \\\\nIntroduction \\\\n One of the key characteristics of superconductivity is a critical temperature (Tc) below this \\\\ntemperature a metal can undergo an electronic transition towards a zero resistance state. Room \\\\ntemperature superconductors (RTSCs) have the potential to revolutionize our world. Perhaps the most \\\\nstraightforward way to reach RTSC can be found within the framework of conventional \\\\nsuperconductivity, where the theory is well established. In principle, Tc can be high in a metal displaying \\\\nsuitable parameters: an appropriate high-frequency phonon spectrum and a strong interaction between \\\\nelectrons and phonons at the Fermi level. The formula derived from the Bardeen-Cooper-Schrieffer and \\\\nMigdal-Eliashberg theories put no apparent limit in Tc 7. Tc, however, depends on a number of factors, \\\\nsuch as the details of the phonon and electronic spectra, which are difficult to estimate. Therefore, the \\\\ntheory provides only a general suggestion for the search of particular systems. Such systems should have \\\\nhigh characteristic phonon frequencies; for example, it can be realized with light atoms and strong \\\\ncovalent bonding. Based on these criteria, superconductivity was found in MgB28 but with a modest Tc of \\\\n39 K. The only material which has been always considered as a potential room-temperature \\\\nsuperconductor is metallic hydrogen9. Although superconductivity in metallic hydrogen has not yet been \\\\nexperimentally observed, this challenge has stimulated the idea to explore other hydrogen containing \\\\ncompounds such as CH4, SiH4, etc., as plausible candidates for the high Tcs.10 These compounds are \\\\ninsulators under ambient conditions, but they can be converted into covalently-bonded metals with the aid \\\\nof accessible pressures, that are much lower than those required for the metallization of pure hydrogen. \\\\nWhile high Tcs have not been found in the initially suggested hydrogen-dominant materials10, the \\\\napproach itself has proven to be very successful: superconductivity with Tc=203 K1 was found in \\\\nhydrogen sulfide, despite a modest hydrogen content in this compound. Further theoretical and \\\\nexperimental research in the H3S related family has not revealed any compound with a higher Tc. For \\\\ninstance, hydrogen selenide has a substantially lower Tc of \\\\uf07e105 K11. The discovery of new metal \\\\nsuperhydrides with the so-called clathrate-like structure of hydrogen atoms raised the calculated Tcs close \\\\nto the room temperature or even above it. In these hydrides, such as in the very first predicted CaH6 with a \\\\nTc of 235 K12, hydrogen atoms create a cage around the host calcium atom. Being connected to each other \\\\nin the network through weak covalent bonding, hydrogen atoms only weakly interact with the host metal \\\\natom through ionic bonding. The host atom supplies a charge to the hydrogen system, stabilizes the \\\\ncrystal structure and promotes metallization at a pressure below 200 GPa. The hydrogen network with \\\\nshort H-H distances in the clathrate-like structures is even closer to metallic atomic hydrogen in \\\\ncomparison to H3S, and can be considered as doped atomic hydrogen. The rapidly growing literature on \\\\nthis topic indicates that various transition metal, lanthanide or actinide elements are prone to form such \\\\nsuperhydrides and some of them exhibits superconductivity with much higher calculated critical \\\\ntemperatures4-6,12-15. The experimental evidence for the record Tc of \\\\uf07e250 K in the fcc-LaH10 at 150 GPa2,3 \\\\nhas confirmed the theoretical predictions and inspired experimentalists to synthesize new members of the \\\\nclathrate-like family of hydrides with promising high Tcs. In the present work, we studied the yttrium-\\\\nhydrogen system, which is the most attractive candidate for very high Tcs among all binary metal-\\\\nhydrogen systems theoretically studied so far. According to the calculations, superconducting fcc-YH10 \\\\nshould have extremely high Tc, that is as high as 303 K at 400 GPa5 or 305−326 K at 250 GPa13. In \\\\naddition to YH10, there are other phases predicted to display very high Tcs and to be stable at lower and \\\\nmore accessible pressures: hcp-YH9 with a Tc=253−276 K can be stabilized at P = 200 GPa5, and bcc-\\\\nYH6 with Tc=251−264 K that is stable already at 110 GPa6. In our experiments, we found the \\\\nsuperconducting phases YH6 and YH9 in agreement with the predicted structures5,6, but with Tcs \\\\nsignificantly lower than prediction by ~30 K. \\\\nResults and Discussion \\\\n Under ambient conditions, the yttrium hydride with highest hydrogen content is hcp-YH3. It is a \\\\nblack narrow-bandgap semiconductor with a metallic luster. When compressed in an inert medium, hcp-\\\\nYH3 converts to fcc-YH316. This pressure-induced phase transformation is extended under a wide pressure \\\\nrange of 10−25 GPa17,18. Further increase of pressure causes likely continuous metallization above \\\\n50 GPa19, as evidenced by the disappearance of the Raman spectrum and a significant drop in resistance. \\\\n Fcc-YH3 was predicted to be stable under high pressures up to 197 GPa20. We found that YH3 and \\\\nYD3 retain the fcc structure of the metallic lattice upon compression without medium up to P=180 GPa \\\\n(Samples 13, 15-18, SM Table 1) and do not exhibit the superconductive transition when subjected to \\\\npressures up to 170 GPa upon cooling down to 5 K. For these samples, we observed the appearance of a \\\\nnew fcc phase in addition to fcc-YH(D)3 under P =130–180 GPa. This new phase atom lattice volume is \\\\nsmaller, than that for fcc-YH(D)3 by ~5 Å3/Y, likely indicating reduced hydrogen content. A similar \\\\nphenomenon was reported earlier upon compression of a substoichiometric LaH3-x in an inert medium21, \\\\nwhere it was interpreted as resulting from a disproportionation reaction into hydrogen-rich stoichiometric \\\\ntrihydride and a hydrogen-poor solid solution. Given that our initial yttrium hydride samples were also \\\\nsubstoichiometric, the appearance of the hydrogen-depleted fcc phase in dry compression runs could also \\\\nresult from a disproportionation reaction. \\\\n When we compressed Y+H2 and Y+D2 mixtures, we observed the formation of YH3 and YD3 already \\\\nat 17 GPa (e.g. in sample 8), judging from a comparison between the atomic volume measured by XRD \\\\nand the literature data for YH318. In such experiments, we did not observe the hydrogen-depleted fcc \\\\nphase with a smaller unit cell parameter, which indicates a complete chemical reaction into the saturated \\\\nstoichiometric yttrium trihydride under excess hydrogen conditions. \\\\n The electrical properties and the structures of the yttrium hydrides with a H stoichiometry higher \\\\nthan YH(D)3 were of particular interest in the present study. Such hydrides were synthesized directly from \\\\nmixtures of Y (samples 3−5), YH3 (samples 1 and 2) or YD3 (samples 6 and 7) in excess of hydrogen \\\\n(deuterium) under high pressures. The chemical reaction occurs already at room temperature after several \\\\nweeks, but can be significantly accelerated with a pulsed laser that can heat the mixture up to \\\\uf07e2000 K. In \\\\nparticular, the I4/mmm YH4 and Im-3m YH6 were synthesized at a pressure of \\\\uf07e160 GPa with the aid of \\\\npulsed laser heating up to \\\\uf07e1000 K (samples 4 and 5 in SM Table 1). After several weeks under higher \\\\npressures between \\\\uf07e200 and 237 GPa, YH4 and YH6 can be synthesized already at room temperature \\\\n(samples 1 and 2 in SM Table 1). The P63/mmc YH9 phase can be synthesized starting from P\\\\uf07e184 GPa \\\\nbut only with the aid of pulsed laser heating (samples 1, 2 and 3, see details in SM Table 1). Higher \\\\npressures apparently promote phase transformation: YH9 is synthesized at 237 GPa even upon a subtle \\\\nsample heating kept below 700 K (no glowing observed) for sample 1 (see the structural details in Fig.2). \\\\n Samples prepared from YH3+H2 mixtures show much sharper transitions and a perfect zero \\\\nresistance state (see the details in Fig.1 (a)). For example, sample 1, which corresponds to the Im-3m \\\\nphase of YH6, was compressed to 237 GPa and kept at room temperature for 3 weeks, showed a sharp \\\\ntransition at 227 K to a zero resistance state (blue curve in Fig.1 (a)).For structural determination see \\\\nFigs.2 (d) and 2 (f). After keeping sample 2 under 201 GPa at room temperature for ~1 month, a Tc of 210 \\\\nK in the Im-3m YH6 phase was observed. With the aid of pulsed laser heating, the Im-3m YH6 phase was \\\\ntransformed into the P63/mmc YH9 phase with a Tc of 243 K (red curve in Fig. 1 (a), see Figs.2 (b) and 2 \\\\n(g) for the details concerning the identification of the structures). The superconductivity with a Tc of 243 \\\\nK could be ascribed to YH9, as samples 1 and 2 displayed Tc values around 210−220 K before the pulsed \\\\nlaser heating (or in the absence of the YH9 phase), and after the pulsed laser heating, the P63/mmc YH9 \\\\nphase was observed with Tc increasing to ~240 K. The main impurity phase in samples 1 and 2 before \\\\nlaser heating was the I4/mmm YH4 phase with c/a ≈ 1.9. This I4/mmm YH4 phase was found in many of \\\\nour samples (1−5), and its XRD pattern is shown in Fig.2 (a). Presently, we are unable to produce \\\\nreasonably pure I4/mmm YH4 to study its superconducting properties. According to calculations6, the \\\\nI4/mmm YH4 phase is superconducting with a Tc=84−95 K, which is considerably lower than those \\\\nmeasured in our samples and Tcs in the range 251−264 K, predicted for bcc-YH6. Thus, superconductivity \\\\nwith a Tc of 227 K and 210 K, as observed in samples 1 and 2 respectively, could be ascribed to the bcc-\\\\nYH6 phase. \\\\n The pressure dependence of Tc for the P63/mmc YH9 and the Im-3m YH6 phases from different \\\\nsamples is summarized in Figs.1 (b) and 1 (c). It is clearly seen in Fig.1 (b) that the pressure dependence \\\\nof Tc for YH9 has a “dome like shape” with the highest Tc at \\\\uf07e243 K under 201 GPa, which is similar to \\\\nthe value found for fcc-LaH102. The range of stability of the YH9 phase differs from the prediction5 – this \\\\nphase is stable at lower pressures. We found that the P63/mmc YH9 phase with higher Tc becomes stable \\\\nfrom at least 185 GPa. The unexpected abrupt drop in Tc(P) in the pressure range of 170−185 GPa, as \\\\nshown by the open black circles in Fig.1 (b), is probably related to the continuous distortion in the crystal \\\\nlattice as observed in SH3 at pressures below \\\\uf07e150 GPa1. In Fig.1 (c), for samples 1 and 2 (mixture of Im-\\\\n3m YH6 and I4/mmm YH4 phases), the onset Tcs were ascribed to the Im-3m YH6 phase. For sample 2, we \\\\ndefined the main superconducting transition at 210 K. A small drop in resistance was also observed at 220 \\\\nK, which is indicated by the smaller open black squares in Fig. 1 (c). However, sample 4 demonstrated a \\\\nhigher Tc~220 K with respect to sample 2. In addition to the Im-3m YH6 and the I4/mmm YH4 phases, \\\\nsample 4 also contained an unidentified complex phase (or a mixture of phases). Because the crystal \\\\nstructure and the stoichiometry of the impurities are not determined, it is not clear whether the \\\\nsuperconductivity observed in sample 4 is attributable to the YH6 phase. Recently, Troyan et al.22 \\\\nobserved superconductivity in the yttrium hydrides, synthesized through laser heating yttrium and \\\\nammonia borane under high pressures. Similarly to sample 4, their samples, revealed a Tc of 224 and 218 \\\\nK at 165 GPa, and comprised a complex mixture of phases; including the bcc-YH6, two new phases with \\\\nclaimed compositions YH7, Y2H15 and an unidentified impurity phase. On the basis of ab-initio \\\\ncalculations, Troyan et al. concluded that these phases should have lower Tcs, and assigned Tc ~220 K to \\\\nbcc-YH6. However, the poor agreement between the experimentally observed XRD patterns and the \\\\nproposed structural models (YH7 and Y2H15 phases) raises concerns about the reliability of their \\\\ninterpretation. Therefore, the superconducting properties of the pure YH6 phase at 160−200 GPa remain \\\\nopen. \\\\n Besides the observed drops in resistance to zero value upon cooling, superconductivity was verified \\\\nby the application of external magnetic fields up to \\\\uf06d0H = 9 T. Figure 1 (d) illustrates the dependence of \\\\nthe superconducting transition from sample 5 on an external magnetic field. Upper critical fields as a \\\\nfunction of temperature following the criterium of 90% of the resistance in the metallic state are shown in \\\\nthe inset of Fig. 1(d). The solid curve in the inset is an extrapolation to estimate the upper critical fields in \\\\nthe limit of zero-temperatures, after fitting the experimental data to the empirical relation: 𝐻𝑐2(𝑇) =\\\\n𝐻𝑐2(0) (1 − (\\\\n𝑇\\\\n𝑇𝑐\\\\n⁄ )\\\\n2\\\\n). This extrapolation yields Hc2(0) =110 T which is about 20 T larger than Hc2(0) \\\\nvalue in H3S 23. The zero resistance for the phases YH6 and YH9 (Fig.1 (a)) as well as the characteristic \\\\nshift of Tc as a function of the magnetic field (Fig.1 (d)) is a clear indication for superconductivity. \\\\n To determine the superconducting pairing mechanism, we substituted hydrogen with deuterium to \\\\nevaluate the effect on Tc. We observed a superconductivity with a Tc of 170 K in sample 6 which was \\\\nsynthesized through the pulsed laser heating of YD3 under deuterium atmosphere at temperatures >1000 \\\\nK and a pressure of 170 GPa. The structural determination still is in progress. \\\\n The crystallographic structure of all samples exhibiting superconducting transitions was determined \\\\nwith the aid of powder X-ray diffraction. The Rietveld refinement for the I4/mmm YH4, Im-3m YH6, and \\\\nP63/mmc YH9 crystal structures are shown in Figs. 2 (a), 2 (b), 2 (c), respectively. Figures 2 (d) and 2 (e) \\\\nare cake representations of the raw X-ray diffraction patterns collected for sample 1 before and after the \\\\npulsed laser heating. Figs. 2 (f) and 2 (g) demonstrate the changes in the powder X-ray diffraction \\\\npatterns during the heating of the mixture of YH3+H2 pressurized at \\\\uf07e200 and \\\\uf07e237 GPa for samples 2 \\\\nand 1, respectively. Pulsed laser heating initiates the phase transformation of YH4 into the YH6 phase, \\\\nand a subsequent transformation of YH6 into the YH9 phase at higher temperatures. \\\\n In order to estimate the stoichiometry of the newly synthesized yttrium hydrides, we studied YH3 in \\\\na wide range of pressures. The experimentally obtained compressibility for the crystal structure of fcc-\\\\nYH3/YD3 phase perfectly coincides with the theoretically calculated data14,20 (Fig.3 (a)). Taking into \\\\naccount the volume occupied by a Y atom in its pure metallic phase at the same pressure24, the additional \\\\nvolume caused by hydration, i.e. the difference between the volumes of YH3 and Y, is 6.7 Å3 at 150 GPa \\\\nand 5.1 Å3 at 215 GPa. Thus, the volume occupied by one hydrogen atom (VH) depends on pressure and \\\\nvaries from 2.2 Å3 to 1.7 Å3 for pressures ranging between 150 and 215 GPa. These estimated values for \\\\nVH are comparable to the one for the La-H system (1.9 Å3 at 150−178 GPa)2 and other metal-H systems25-\\\\n27. Using the calculated values of 1.6 Å3 for VH and 11.2 Å3 for the volume of yttrium derived from the \\\\nextrapolated data from the equation of state of metallic yttrium24, the stoichiometry calculated from the \\\\nexperimental diffraction data for the new yttrium hydrides are YH4.1, YH5.8, and YH8.8 at 237 GPa, \\\\nrespectively. The crystallographic structures of YH4, YH6, and YH9, agree perfectly with the theoretical \\\\npredictions5,6. The X-ray diffraction data for the volume of the crystal structure normalized with respect to \\\\nthe volume one Y atom as well as fragments of the crystal structures and coordination polyhedra for all \\\\nexperimentally found yttrium hydride phases are shown in Figure 3. The positions of the yttrium atoms \\\\nwere found directly from the diffraction data, whereas the hydrogen atoms were placed in the \\\\ngeometrically calculated positions based on the theoretical data5. \\\\n In spite of the very good agreement between the predictions and the experimental crystallographic \\\\nstructures, the measured Tcs for the Im-3m YH6 (Tc \\\\uf07e227 K) and the P63/mmc YH9 (Tc \\\\uf07e243 K) phases are \\\\nmarkedly lower than the predicted values of 251−264 K for YH66 and 253−276 K for YH95. Thus, only \\\\nthe fcc-YH10 phase can be expected to display RTSC with a predicted Tc \\\\uf07e 305−326 K13. However, we did \\\\nnot find YH10, in spite of extensive trials with high-pressure synthesis up to 237 GPa. Still there is a \\\\npossibility that this phase can be synthesized at higher pressures and temperatures. According to some \\\\npredictions, the fcc-YH10 phase is dynamically stable starting from 226 GPa15 or 220 GPa13. However, \\\\nother calculations5 suggest that the fcc-YH10 cannot be stabilized even at pressures as high as 400 GPa. \\\\nInstead, the hexagonal YH9 is energetically more favorable and lies on both convex hulls of formation \\\\nenthalpy and internal energy, while YH10 has a higher formation enthalpy and lies above the convex hull \\\\nby 24 meV/atom at 400 GPa that is associated to 1100 K5. The synthesis attempt of fcc-YH10 under \\\\nhigher pressures and temperatures is in progress. \\\\n \\\\nMethods \\\\n To synthesize the initial YH3 and YD3, yttrium metal of 99.9% purity was first annealed in a vacuum \\\\nof about 10-3 Torr at 400 °C for 4 h, and then exposed to hydrogen (or deuterium) gas at a pressure of \\\\nabout 100 bars at 200 °C for 24 h in a high-pressure Sievert-type apparatus. The sample treatment was \\\\ndone in an Ar glovebox to prevent oxidation. The reaction products were YH2.92(5) and YD2.87(5) as \\\\nindicated by their weight. We will further refer to these samples as YH3 and YD3 for brevity. The \\\\nsamples were analyzed through XRD with an Empyrean diffractometer at ambient conditions under a \\\\nKapton film. The lattice parameters of YH3 and YD3 were in reasonable agreement with the available \\\\ndata28. \\\\n In the diamond anvil cells (DACs), we typically synthesized yttrium hydride via a direct reaction \\\\nbetween yttrium (Alfa Aesar 99.9%) or YH3 and hydrogen (99.999%) at high pressures. For that, a piece \\\\nof Y or YH3 was placed into a hole drilled in an insulating gasket. The process of synthesis is the same as \\\\nthe one followed for lanthanum hydride2. The pressure, pulsed laser heating temperature, and the amount \\\\nof hydrogen surrounding the sample determined the composition of the yttrium hydrides. Superhydrides \\\\nwere synthesized only under an evident excess of hydrogen and high enough pressure. For the thermal \\\\ntreatment, one-sided pulsed radiation from a YAG laser was focused onto a spot having a diameter of 10 \\\\nµm. Some samples were prepared not from elemental yttrium as the starting material but from YH3 which \\\\nwas synthesized and analyzed as described above. One of the advantages of this method is to initially \\\\nhave a higher hydrogen content. To determine the isotope effect, we substituted hydrogen by deuterium \\\\n(99.75% purity). \\\\n Typically, the diamonds used in the DACs had a culet with a diameter of 20−35 µm and were \\\\nbeveled at 8° to a diameter of about 250 µm. A toroidal profile was machined at each culet by a focused \\\\nbeam of xenon ions (FERA3 TESCAN). Tantalum electrodes were sputtered onto the surface of one of \\\\nthe diamond anvils in a van der Pauw four-probe geometry and were covered with gold. A metallic gasket \\\\n(T301 stainless steel) was electrically separated from the electrodes by an insulating layer (a mixture of \\\\nepoxy and CaF2, MgO, CaSO4, cBN or Al2O3). The typical sample size was 5−10 µm. \\\\n We present resistance measurements upon warming the DACs as it yields a more accurate \\\\ntemperature reading: the cell was warmed up slowly (0.2 K min-1) under an isothermal environment (no \\\\ncoolant flow). The temperature was measured with an accuracy of about 0.1 K by a Si diode thermometer \\\\nattached to the DAC. All electrical transport measurements were performed with the electrical current set \\\\nin the range of 10-5-10-3 A. The pressure was measured through the H2 (D2) vibron scale29 if such a vibron \\\\ncan be observed, or otherwise from the diamond Raman edge scale30. The Tc was determined from the \\\\nonset of superconductivity – at the point of apparent deviation from the temperature dependence of the \\\\nresistance in the normal state metallic behavior. \\\\n We used three types of DACs. In particular, DACs with diameters of 20 mm and 8.8 mm were made \\\\nof nonmagnetic materials, suitable for measurements under magnetic fields using a 9 T Quantum Design \\\\nPhysical Property Measurement System (PPMS). The X-ray diffraction measurements were done with \\\\nwavelengths of 0.3344 Å and 0.2952 Å, an X-ray spot size ~3×3 µm, and Pilatus 1M CdTe detector at the \\\\nbeamline 13-IDD at GSECARS, Advanced Photon Source, Argonne National Laboratory (Chicago). \\\\nPrimary processing and integration of the powder patterns were made using the Dioptas software31. The \\\\nIndexing and refinement were done with GSAS and EXPGUI packages32. \\\\n \\\\nFigure captions \\\\nFigure 1. Superconducting transitions in yttrium hydrides. \\\\n(a) The temperature dependence of resistance for the Im-3m YH6 phase (blue curve, sample 1) and the \\\\nP63/mmc YH9 phase (red curve, sample 2). Samples 1 and 2 were synthesized from YH3 under hydrogen \\\\natmosphere. The configuration of the measurements is shown in the inset. After 3 weeks of maintaining \\\\nsample 1 at 237 GPa and at room temperature, a sharp transition appeared, indicating a superconducting \\\\ntransition with an onset of Tc= 227 K – blue curve. This transition corresponds to the Im-3m YH6 phase \\\\n(see the details of the structure in Fig.2 (d) and 2 (f)). After keeping sample 2 at 201 GPa and at room \\\\ntemperature for ~ 1 month, with the aid of pulsed laser heating, the P63/mmc YH9 phase was observed \\\\nwith a Tc =243 K (see Fig.2 (b) and 2 (g) for the structural characterization). \\\\n(b) The pressure dependence of Tc for the different samples belonging to the P63/mmc YH9 phase: \\\\nS1−filled magenta circles−sample 1 (after laser heating), S2 –filled black circles – sample 2 (after laser \\\\nheating), S3−filled red circles–sample 3. Details concerning the synthesis of the samples can be found in \\\\nthe SM Table 1. Data shown by open black circles were obtained by decreasing the pressure of sample 2. \\\\nThe open magenta circles correspond to the decompression of sample 1. The dotted line is a guide to the \\\\neyes. \\\\n(c) The pressure dependence of Tc for the different samples belonging to the Im-3m YH6 phase: S1−filled \\\\nmagenta squares−sample 1 (before laser heating), S2−filled black squares with main the superconducting \\\\ntransition at 210 K, smaller open black squares represent a small drop in resistance at 220 K−sample 2 \\\\n(before laser heating), S4−filled olive squares–sample 4. The details concerning sample synthesis can be \\\\nfound in SM Table 1. Samples 1 and 2 crystallize in the Im-3m YH6 and in the I4/mmm YH4 phases, with \\\\nthe Tc dominated by the Im-3m YH6 phase as indicated by the black polygon. Sample 4 crystallized in the \\\\nIm-3m YH6 and I4/mmm YH4 phases in addition to a complex unidentified phase (or a mixture of phases). \\\\n(d) Superconducting transition for sample 5 under an external magnetic field. This panel displays the \\\\nelectrical resistance as a function of the temperature under applied magnetic fields up to 9 T. For sample \\\\n5, R (T) was measured through a three terminal method, and the resulting background was subsequently \\\\nsubtracted. The temperature dependence of the observed upper critical fields (inset) was obtained from \\\\nthe data shown in Fig. 1 (d). An extrapolation to the lowest temperatures yields an \\\\uf07e110 T for the upper \\\\ncritical magnetic field in the limit of zero temperatures. \\\\nFigure 2. X-ray powder diffraction analysis of the synthesized yttrium hydrides through Rietveld \\\\nrefinement of the crystal structures. Black crosses are the experimental points, solid red lines are the \\\\nRietveld fits, and the blue solid lines represent the residuals. The blue, magenta and green ticks indicate \\\\nthe diffraction peaks corresponding to the tetragonal YH4, cubic YH6 and hexagonal YH9 phases, \\\\nrespectively. The used X-ray wavelengths (λ, Å) are indicated in the top right corners of each plot. The \\\\nrelative contribution between phases in a Rietveld refinement comprising two different crystalline phases \\\\nis summarized at the bottom right corner of each plot. Fit parameters for each refinement are shown under \\\\nthe blue residual graph. \\\\n(a) Rietveld refinement for the I4/mmm YH4 crystal structure found as a pure single-phase in some \\\\nregions of sample 4 subjected to 183 GPa. \\\\n(b) Rietveld refinement for the Im-3m YH6 crystal structure found in sample 2 under 201 GPa which was \\\\nkept at room temperature for 1 month. \\\\n(c) Rietveld refinement for the P63/mmc YH9 crystallographic structure found in sample 1 under 237 GPa \\\\nafter annealing for five cycles of very subtle pulsed laser heating (the temperature of each cycle was kept \\\\nbelow 700 K). \\\\n(d) Cake representation of the raw X-ray diffraction patterns collected for sample 1, which was kept at \\\\nroom temperature for 3 weeks. The three broad and highly spotted lines correspond to the Im-3m phase of \\\\nYH6 (this raw pattern corresponds to the black powder pattern in (f)). \\\\n(e) The same cake representation of sample 1 after five cycles of very subtle heating up to temperatures \\\\nbelow 700 K. The new narrow and dashed lines appearing after the heat treatment and correspond to the \\\\nP63/mmc phase of YH9 (this raw pattern corresponds to the green powder pattern in (f)). \\\\n(f) The changes in the powder patterns of sample 1 after successive cycles of very subtle heating (the \\\\ntemperature of each cycle is maintained below 700 K). The black powder pattern before any pulsed laser \\\\nheating corresponds to the nearly pure Im-3m phase of YH6. Each heating cycle results in the conversion \\\\nof YH6 into the higher YH9 hydride, as indicated by the appearance of new diffraction peaks related to the \\\\nP63/mmc lattice. Reflections from the Im-3m YH6 and P63/mmc YH9 phases are marked by c and h, \\\\nrespectively. \\\\n(g) Changes in the powder diffraction patterns after several cycles of pulsed laser heating for sample 2. \\\\nThe laser power and the corresponding temperature at each successive heating cycle was continuously \\\\nincreased up to 1750(15) K. At the beginning, the sample mainly consists of the Im-3m YH6 and I4/mmm \\\\nYH4 phases. Each heating cycle that maintains the temperature below ~1300 K initiates the \\\\ntransformation of the YH4 phase into YH6. Successive heating cycles to higher temperatures initiate the \\\\ntransformation of YH6 into the YH9 phase. \\\\n \\\\nFigure 3. Equation of state and crystal structures for the different YHx (0 ≤ x ≤ 9) phases. \\\\n(a) Unit cell volume normalized to the volume of one Y atom as a function of the pressure for the \\\\ndifferent yttrium hydrides as well as for pure yttrium. Data related to the same given phase are indicated \\\\nby the red polygons. Hexagons, stars, squares, triangles, and filled circles correspond to P63/mmc YH9, \\\\nIm-3m YH6, I4/mmm YH4, Fm-3m YH3, and distorted Fm-3m (hR24 and C2/m) yttrium, respectively. \\\\nRed, blue, green, magenta, purple, grey, navy, orange, azure, dark green, light grey, dark yellow, brown, \\\\ndark violet, black, light pink, light brown markers correspond to samples 1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, \\\\n13, 14, 15, 16, 17, and 18, respectively (see the SM Table 1). Filled black circles correspond to the \\\\nexperimental data for metallic yttrium24. Open black symbols correspond to the theoretically predicted \\\\ncrystallographic structures: open stars for YH65,6, open squares for YH46, and open triangles for YH314,20. \\\\n(b) Crystallographic structure for the different yttrium hydrides found in the experiment based on the X-\\\\nray diffraction data from the lattice of the heavier yttrium atoms. Hydrogen atoms were geometrically \\\\nplaced using their positions according to the theoretically predicted structures5. Yttrium and hydrogen \\\\natoms are coloured in blue and grey, respectively. The unit cell is indicated by red lines. The coordination \\\\npolyhedra (cages) and building polygons for these structures (YH3, YH4, YH6, YH9) are indicated by the \\\\nlight gray lines with the corresponding compositions shown in the row immediately below each structure. \\\\n \\\\n \\\\nAcknowledgements. \\\\nM.E. is thankful to the Max Planck community for the invaluable support, and U. Pöschl for the constant \\\\nencouragement. L.B. is supported by DOE−BES through award DE-SC0002613. S.M. acknowledges \\\\nsupport from the FSU Provost Postdoctoral Fellowship Program. The NHMFL acknowledges support from \\\\nthe U.S. NSF Cooperative Grant No. DMR−1644779, and the State of Florida. Portions of this work were \\\\nperformed at GeoSoilEnviro CARS (The University of Chicago, Sector 13), Advanced Photon Source \\\\n(APS), Argonne National Laboratory. GeoSoilEnviro CARS is supported by the National Science \\\\nFoundation−Earth Sciences (EAR−1634415) and Department of Energy−GeoSciences \\\\n(DE−FG02−94ER14466). This research used resources of the Advanced Photon Source, a U.S. Department \\\\nof Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne \\\\nNational Laboratory under Contract No. DE−AC02−06CH11357. \\\\n \\\\nCorrespondence and requests for materials should be addressed to M.E. \\\\n \\\\nReferences \\\\n1 Drozdov, A. P., Eremets, M. I., Troyan, I. A., Ksenofontov, V. & Shylin, S. I. Conventional \\\\nsuperconductivity at 203 kelvin at high pressures in the sulfur hydride system. Nature 525, 73 \\\\n(2015). \\\\n2 Drozdov, A. P. et al. Superconductivity at 250 K in lanthanum hydride under high pressures. \\\\nNature 569, 528 (2019). \\\\n3 Somayazulu, M. et al. Evidence for Superconductivity above 260 K in Lanthanum Superhydride \\\\nat Megabar Pressures. Phys Rev Lett 122, 027001 (2019). \\\\n4 Liu, H. et al. Dynamics and superconductivity in compressed lanthanum superhydride. Physical \\\\nReview B 98, 100102 (2018). \\\\n5 Peng, F. et al. Hydrogen Clathrate Structures in Rare Earth Hydrides at High Pressures: Possible \\\\nRoute to Room-Temperature Superconductivity. Phys Rev Lett 119, 107001 (2017). \\\\n6 Li, Y. et al. Pressure-stabilized superconductive yttrium hydrides. Sci Rep 5, 9948 (2015). \\\\n7 Allen, P. B. & Dynes, R. C. Transition temperature of strong-coupled superconductors \\\\nreanalyzed. Physical Review B 12, 905 (1975). \\\\n8 Nagamatsu, J., Nakagawa, N., Muranaka, T., Zenitani, Y. & Akimitsu, J. Superconductivity at 39 K \\\\nin magnesium diboride. Nature 410, 63 (2001). \\\\n9 Ashcroft, N. W. Metallic Hydrogen: A High-Temperature Superconductor? Physical Review \\\\nLetters 21, 1748 (1968). \\\\n10 Ashcroft, N. W. Hydrogen dominant metallic alloys: high temperature superconductors? Phys \\\\nRev Lett 92, 187002 (2004). \\\\n11 Mishra, A. et al. Novel Synthesis Route and Observation of Superconductivity in the Se-H System \\\\nat Extreme Conditions. APS March Meeting 2018 (2018). \\\\n12 Wang, H., Tse, J. S., Tanaka, K., Iitaka, T. & Ma, Y. Superconductive sodalite-like clathrate calcium \\\\nhydride at high pressures. Proc. Natl. Acad. Sci. 109, 6463 (2012). \\\\n13 Liu, H., Naumov, II, Hoffmann, R., Ashcroft, N. W. & Hemley, R. J. Potential high-Tc \\\\nsuperconducting lanthanum and yttrium hydrides at high pressure. Proc. Natl. Acad. Sci. 114, \\\\n6990 (2017). \\\\n14 Liu, L. L., Sun, H. J., Wang, C. Z. & Lu, W. C. High-pressure structures of yttrium hydrides. J Phys \\\\nCondens Matter 29, 325401 (2017). \\\\n15 Heil, C., di Cataldo, S., Bachelet, G. B. & Boeri, L. Superconductivity in sodalite-like yttrium \\\\nhydride clathrates. Physical Review B 99, 220502 (2019). \\\\n16 Palasyuk, T. & Tkacz, M. Hexagonal to cubic phase transition in YH3 under high pressure. Solid \\\\nState Communications 133, 477 (2005). \\\\n17 Machida, A., Ohmura, A., Watanuki, T., Aoki, K. & Takemura, K. Long-period stacking structures \\\\nin yttrium trihydride at high pressure. Physical Review B 76, 052101 (2007). \\\\n18 Machida, A. et al. X-ray diffraction investigation of the hexagonal–fcc structural transition in \\\\nyttrium trihydride under hydrostatic pressure. Solid State Communications 138, 436-440 (2006). \\\\n19 Nguyen, H., Chi, Z., Matsuoka, T., Kagayama, T. & K., S. Pressure-induced metallization of yttrium \\\\ntrihydride, YH3. J. Phys. Soc. Jpn. 81, SB041 (2012). \\\\n20 Li, Y. & Ma, Y. Crystal structures of YH3 under high pressure. Solid State Communications 151, \\\\n388 (2011). \\\\n21 Machida, A., Watanuki, T., Kawana, D. & Aoki, K. Phase separation of lanthanum hydride under \\\\nhigh pressure. Physical Review B 83, 054103 (2011). \\\\n22 Troyan, I. A. et al. Synthesis and Superconductivity of Yttrium Hexahydride Im-3m-YH6. arXiv e-\\\\nprints (2019). <https://ui.adsabs.harvard.edu/abs/2019arXiv190801534T>. \\\\n23 Mozaffari, S. et al. Superconducting phase diagram of H3S under high magnetic fields. Nature \\\\nCommunications 10, 2522 (2019). \\\\n24 Samudrala, G. K., Tsoi, G. M. & Vohra, Y. K. Structural phase transitions in yttrium under \\\\nultrahigh pressures. Journal of Physics: Condensed Matter 24, 362201 (2012). \\\\n25 Scheler, T. et al. High-pressure synthesis and characterization of iridium trihydride. Phys Rev Lett \\\\n111, 215503 (2013). \\\\n26 Pépin, C. M., Dewaele, A., Geneste, G. & Loubeyre, P. New Iron Hydrides under High Pressure. \\\\nPhys. Rev. Lett. 113, 265504 (2014). \\\\n27 Pépin, C. M., Geneste, G., Dewaele, A., Mezouar, M. & Loubeyre, P. Synthesis of FeH5: A layered \\\\nstructure with atomic hydrogen slabs. Science 357, 382 (2017). \\\\n28 Fedotov, V. K., Antonov, V. E., Bashkin, I. O., Hansen, T. & Natkaniec, I. Displacive ordering in the \\\\nhydrogen sublattice of yttrium trihydride. Journal of Physics: Condensed Matter 18, 1593 (2006). \\\\n29 Eremets, M. I. & Troyan, I. A. Conductive dense hydrogen. Nat Mater 10, 927 (2011). \\\\n30 Eremets, M. I. Megabar high-pressure cells for Raman measurements. Journal of Raman \\\\nSpectroscopy 34, 515 (2003). \\\\n31 Prescher, C. & Prakapenka, V. B. DIOPTAS: a program for reduction of two-dimensional X-ray \\\\ndiffraction data and data exploration. High Pressure Research 35, 223 (2015). \\\\n32 Toby, B. H. EXPGUI, a graphical user interface for GSAS. Journal of applied crystallography 34, \\\\n210 (2001). \\\\n \\\\nFigure 1. Superconducting transitions in yttrium hydrides. \\\\n(a) The temperature dependence of resistance for the Im-3m YH6 phase (blue curve, sample 1) and the \\\\nP63/mmc YH9 phase (red curve, sample 2). Samples 1 and 2 were synthesized from YH3 under hydrogen \\\\natmosphere. The configuration of the measurements is shown in the inset. After 3 weeks of maintaining \\\\nsample 1 at 237 GPa and at room temperature, a sharp transition appeared, indicating a \\\\nsuperconducting transition with an onset of Tc= 227 K – blue curve. This transition corresponds to the \\\\nIm-3m YH6 phase (see the details of the structure in Fig.2 (d) and 2 (f)). After keeping sample 2 at 201 \\\\nGPa and at room temperature for ~ 1 month, with the aid of pulsed laser heating, the P63/mmc YH9\\\\nphase was observed with a Tc =243 K (see Fig.2 (b) and 2 (g) for the structural characterization).\\\\n(b) The pressure dependence of Tc for the different samples belonging to the P63/mmc YH9 phase: \\\\nS1−filled magenta circles−sample 1 (after laser heating), S2 –filled black circles – sample 2 (after laser \\\\nheating), S3−filled red circles–sample 3. Details concerning the synthesis of the samples can be found in \\\\nthe SM Table 1. Data shown by open black circles were obtained by decreasing the pressure of sample 2. \\\\nThe open magenta circles correspond to the decompression of sample 1. The dotted line is a guide to \\\\nthe eyes. \\\\n(c) The pressure dependence of Tc for the different samples belonging to the Im-3m YH6 phase: S1−filled \\\\nmagenta squares−sample 1 (before laser heating), S2−filled black squares with main the \\\\nsuperconducting transition at 210 K, smaller open black squares represent a small drop in resistance at \\\\n220 K−sample 2 (before laser heating), S4−filled olive squares–sample 4. The details concerning sample \\\\nsynthesis can be found in SM Table 1. Samples 1 and 2 crystallize in the Im-3m YH6 and in the I4/mmm\\\\nYH4 phases, with the Tc dominated by the Im-3m YH6 phase as indicated by the black polygon. Sample 4 \\\\ncrystallized in the Im-3m YH6 and I4/mmm YH4 phases in addition to a complex unidentified phase (or a \\\\nmixture of phases). \\\\n(d) Superconducting transition for sample 5 under an external magnetic field. This panel displays the \\\\nelectrical resistance as a function of the temperature under applied magnetic fields up to 9 T. For sample \\\\n5, R (T) was measured through a three terminal method, and the resulting background was subsequently \\\\nsubtracted. The temperature dependence of the observed upper critical fields (inset) was obtained from \\\\nthe data shown in Fig. 1 (d). An extrapolation to the lowest temperatures yields an \\\\uf07e110 T for the upper \\\\ncritical magnetic field in the limit of zero temperatures. \\\\n150 180 210 240\\\\n220\\\\n230\\\\n240\\\\n250\\\\nPressure, GPa\\\\nYH\\\\n9\\\\nT\\\\nc\\\\n, \\\\nK\\\\n(b)\\\\nS1\\\\nS2\\\\nS3\\\\n0 50 100 150 200 250\\\\n0\\\\n20\\\\n40\\\\n60\\\\n80\\\\n100\\\\n120\\\\n150 200 250 300\\\\n0\\\\n20\\\\n40\\\\n60\\\\n80\\\\n\\\\uf06d\\\\n\\\\uf030\\\\nH\\\\n, \\\\nT\\\\nT , K\\\\nR\\\\ne\\\\ns\\\\nis\\\\nta\\\\nn\\\\nc\\\\ne\\\\n, \\\\n\\\\uf057\\\\nTemperature, K\\\\n 0 T\\\\n 1 T\\\\n 3 T\\\\n 5 T\\\\n 7 T\\\\n 9 T\\\\n(d)\\\\n150 200 250\\\\n200\\\\n220\\\\n240\\\\nPressure, GPa\\\\npure YH\\\\n6\\\\n+YH\\\\n4\\\\n phases\\\\nT\\\\nc\\\\n, \\\\nK\\\\n(c)\\\\nYH\\\\n6\\\\n+YH\\\\n4\\\\n+unidentified impurity\\\\nS2\\\\nS1\\\\nS4\\\\nS2\\\\n50 100 150 200 250 300\\\\n0.0\\\\n0.2\\\\n0.4\\\\n0.6\\\\n \\\\n \\\\nR*12\\\\nYH\\\\n6\\\\n T\\\\nc\\\\n = 227 K\\\\n P = 237 GPa\\\\nYH\\\\n9\\\\n T\\\\nc\\\\n = 243 K\\\\n P = 201 GPa\\\\nR\\\\ne\\\\ns\\\\nis\\\\nta\\\\nn\\\\nc\\\\ne\\\\n, \\\\n\\\\uf057\\\\nTemperature, K(a)\\\\nYH3 sample H2 electrode\\\\nFigure 2. X-ray powder diffraction analysis of the synthesized yttrium hydrides through Rietveld \\\\nrefinement of the crystal structures. Black crosses are the experimental points, solid red lines are the \\\\nRietveld fits, and the blue solid lines represent the residuals. The blue, magenta and green ticks \\\\nindicate the diffraction peaks corresponding to the tetragonal YH4, cubic YH6 and hexagonal YH9\\\\nphases, respectively. The used X-ray wavelengths (λ, Å) are indicated in the top right corners of each \\\\nplot. The relative contribution between phases in a Rietveld refinement comprising two different \\\\ncrystalline phases is summarized at the bottom right corner of each plot. Fit parameters for each \\\\nrefinement are shown under the blue residual graph. \\\\n(a) Rietveld refinement for the I4/mmm YH4 crystal structure found as a pure single-phase in some \\\\nregions of sample 4 subjected to 183 GPa. \\\\n(b) Rietveld refinement for the Im-3m YH6 crystal structure found in sample 2 under 201 GPa which was \\\\nkept at room temperature for 1 month.\\\\n(c) Rietveld refinement for the P63/mmc YH9 crystallographic structure found in sample 1 under 237 \\\\nGPa after annealing for five cycles of very subtle pulsed laser heating (the temperature of each cycle \\\\nwas kept below 700 K).\\\\n(d) Cake representation of the raw X-ray diffraction patterns collected for sample 1, which was kept at \\\\nroom temperature for 3 weeks. The three broad and highly spotted lines correspond to the Im-3m\\\\nphase of YH6 (this raw pattern corresponds to the black powder pattern in (f)).\\\\n(e) The same cake representation of sample 1 after five cycles of very subtle heating up to \\\\ntemperatures below 700 K. The new narrow and dashed lines appearing after the heat treatment and \\\\ncorrespond to the P63/mmc phase of YH9 (this raw pattern corresponds to the green powder pattern in \\\\n(f)).\\\\n(f) The changes in the powder patterns of sample 1 after successive cycles of very subtle heating (the \\\\ntemperature of each cycle is maintained below 700 K). The black powder pattern before any pulsed \\\\nlaser heating corresponds to the nearly pure Im-3m phase of YH6. Each heating cycle results in the \\\\nconversion of YH6 into the higher YH9 hydride, as indicated by the appearance of new diffraction peaks \\\\nrelated to the P63/mmc lattice. Reflections from the Im-3m YH6 and P63/mmc YH9 phases are marked \\\\nby c and h, respectively.\\\\n(g) Changes in the powder diffraction patterns after several cycles of pulsed laser heating for sample 2. \\\\nThe laser power and the corresponding temperature at each successive heating cycle was continuously \\\\nincreased up to 1750(15) K. At the beginning, the sample mainly consists of the Im-3m YH6 and \\\\nI4/mmm YH4 phases. Each heating cycle that maintains the temperature below ~1300 K initiates the \\\\ntransformation of the YH4 phase into YH6. Successive heating cycles to higher temperatures initiate the \\\\ntransformation of YH6 into the YH9 phase.\\\\n4 6 8 10 12 14 16 18 20\\\\n4 6 8 10 12 14 16 18 20\\\\n-1\\\\n0\\\\n1\\\\n2\\\\n3\\\\n4 6 8 10 12 14 16 18 20 22\\\\n4 6 8 10 12 14 16 18 20 22\\\\n-3\\\\n0\\\\n3\\\\n6\\\\n9\\\\n12\\\\n4 5 6 7 8 9 10 11 12 13 14\\\\n4 5 6 7 8 9 10 11 12 13 14\\\\n0\\\\n1\\\\n2\\\\n(c)\\\\n(b)\\\\nY 2d 0.333 0.667 0.750\\\\nY 2a 0 0 0\\\\na = 3.364(1) Å\\\\nc = 5.153(1)Å\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\n Iexp\\\\n Icalc\\\\n Iexp- Icalc\\\\n I4/mmm YH4\\\\nRp=19.0%, wR=37.9%\\\\nl=0.3344 Å\\\\na = 2.708(1) Å\\\\nc = 5.198(2) Å\\\\nY 2a 0 0 0\\\\n(a)\\\\na = 3.492(1) Å\\\\nl=0.2952 Å\\\\n Im-3m YH6\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\nRp=20.1%, wR=47.4%\\\\ntI:cI = 0.16 : 1\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\n P63/mmc YH9l=0.3344Å\\\\nRp=21.1%, wR=35.4%\\\\ncI:hP = 0.93 : 1\\\\n8 10 12 14\\\\n0\\\\n3\\\\n6\\\\n9\\\\n12\\\\n15\\\\n18 l=0.3344 Å\\\\n5\\\\n3\\\\n1\\\\nh\\\\nh\\\\nh h h\\\\nc\\\\nc\\\\nc\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\nN\\\\no\\\\n. \\\\no\\\\nf \\\\nh\\\\ne\\\\na\\\\nti\\\\nn\\\\ng\\\\n c\\\\ny\\\\nc\\\\nle\\\\n0\\\\n(f)\\\\n7 8 9 10 11 12\\\\n0\\\\n1\\\\n2\\\\n3\\\\n4\\\\n5\\\\nt\\\\nl=0.2952 Å\\\\nh\\\\nh\\\\nh h\\\\nc\\\\nc\\\\nc\\\\nt t tIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\n\\\\uf07e300\\\\n\\\\uf07e650\\\\n\\\\uf07e1050\\\\n\\\\uf07e1300\\\\n\\\\uf07e1750\\\\nT, K\\\\n(g)\\\\n6 8 10 12 14\\\\n-180\\\\n-120\\\\n-60\\\\n0\\\\n60\\\\n120\\\\n180\\\\n6 8 10 12 14\\\\n-180\\\\n-120\\\\n-60\\\\n0\\\\n60\\\\n120\\\\n180\\\\n2q, °\\\\nA\\\\nz\\\\nim\\\\nu\\\\nth\\\\n, \\\\n°\\\\nA\\\\nz\\\\nim\\\\nu\\\\nth\\\\n, \\\\n°\\\\n2q, °\\\\n(d)\\\\n(e)\\\\nYH3\\\\nYH14\\\\n12 H4\\\\nYH4\\\\nYH18\\\\n4 H6 + 8 H4\\\\nYH6\\\\nYH24\\\\n6 H6 + 6 H4\\\\nYH9\\\\nYH29\\\\n6 H6 + 6 H5 + 6 H4\\\\n(b)\\\\n50 100 150 200 250\\\\n12\\\\n14\\\\n16\\\\n18\\\\n20\\\\n22\\\\n24\\\\n26\\\\n28\\\\nYH\\\\n4\\\\nYH\\\\n9\\\\nV\\\\no\\\\nlu\\\\nm\\\\ne\\\\n p\\\\ne\\\\nr \\\\na\\\\nto\\\\nm\\\\n, \\\\nÅ\\\\n3\\\\nPressure, GPa\\\\nY\\\\nYH\\\\n3\\\\nYH\\\\n6\\\\nhR24\\\\nC2/m\\\\n(a)\\\\nFigure 3. Equation of states and crystal structures of different YHx (0 ≤ x ≤ 9) phases.\\\\n(a) Unit cell volume normalized to the volume of one Y atom as a function of the pressure for the different \\\\nyttrium hydrides as well as for pure yttrium. Data related to the same given phase are indicated by the \\\\nred polygons. Hexagons, stars, squares, triangles, and filled circles correspond to P63/mmc YH9, Im-3m\\\\nYH6, I4/mmm YH4, Fm-3m YH3, and distorted Fm-3m (hR24 and C2/m) yttrium, respectively. Red, blue, \\\\ngreen, magenta, purple, grey, navy, orange, azure, dark green, light grey, dark yellow, brown, dark violet, \\\\nblack, light pink, light brown markers correspond to samples 1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, \\\\n17, and 18, respectively (see the SM Table 1). Filled black circles correspond to the experimental data for \\\\nmetallic yttrium24. Open black symbols correspond to the theoretically predicted crystallographic \\\\nstructures: open stars for YH6\\\\n5,6, open squares for YH4\\\\n6, and open triangles for YH3\\\\n14,20.\\\\n(b) Crystallographic structure for the different yttrium hydrides found in the experiment based on the X-\\\\nray diffraction data from the lattice of the heavier yttrium atoms. Hydrogen atoms were geometrically \\\\nplaced using their positions according to the theoretically predicted structures5. Yttrium and hydrogen \\\\natoms are coloured in blue and grey, respectively. The unit cell is indicated by red lines. The coordination \\\\npolyhedra (cages) and building polygons for these structures (YH3, YH4, YH6, YH9) are indicated by the light \\\\ngray lines with the corresponding compositions shown in the row immediately below each structure.\\\\nSample Synthesis conditions \\\\nElectrical \\\\nmeasurements \\\\nX-ray diffraction \\\\n1 (1010) \\\\nYH3 + H2. Pressurized to 255 (or 237 \\\\naccording to H2 vibron) GPa and kept for \\\\n3 weeks at room temperature. \\\\nTc = 227 K \\\\nFig.1c, magenta \\\\nsquare \\\\nIm-3m YH6, dominant phase: \\\\na=3.457(1) Å, V=41.3(1) Å3; \\\\nI4/mmm YH4: \\\\na=2.616 Å, c=5.184 Å, V=35.5 Å3. \\\\nThe same sample after 5 subtle laser \\\\nheating cycles (the temperature of each \\\\ncycle is below 700 K, no visible glowing). \\\\nTc = 237 K \\\\nFig.1b, filled \\\\nmagenta circle \\\\nP63/mmc YH9 phase: \\\\na=3.364(1) Å, c=5.153(1) Å, V=50.5(1) \\\\nÅ3; \\\\nwith remains of Im-3m YH6 and \\\\nI4/mmm YH4. \\\\n2 (Play) \\\\nYH3 + H2. Pressurized to 238 (or 201 \\\\naccording to H2 vibron) GPa and kept for \\\\n1 months at room temperature. \\\\nFig.1a, blue \\\\ncurve, Fig.1c, \\\\nblack square \\\\n Tc = 210 K \\\\nIm-3m YH6, dominant phase: \\\\na=3.492(1) Å, V=42.6(1) Å3; \\\\nwith I4/mmm YH4: \\\\na=2.656(1) Å, c=5.190(8) Å, V=36.6(1) \\\\nÅ3. \\\\nThe same sample after tens of laser \\\\nheating cycles up to 2000(10) K. \\\\nFig.1a, red \\\\ncurve, Fig.1b, \\\\nfilled black circle \\\\nTc = 243 K \\\\nP63/mmc YH9 phase: \\\\na=3.406 Å, c=5.210 Å, V=52.3 Å3; \\\\nwith remains of Im-3m YH6 and traces \\\\nof I4/mmm YH4. \\\\n3 (H120) \\\\nY + H2. Pressurized to 235 (or 184 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nFig.1b, filled red \\\\ncircle \\\\n Tc = 239 K \\\\nI4/mmm YH4, dominant phase: \\\\na=2.666 Å, c=5.194 Å, V=36.9 Å3; \\\\nIm-3m YH6 phase: \\\\na=3.529 Å, V=43.9 Å3; \\\\nP63/mmc YH9 phase: \\\\na=3.432 Å, c=5.251 Å, V=53.6 Å3. \\\\n4 (YH1) \\\\nY + H2. Pressurized to 189 (or 183 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nFig.1c, olive \\\\nsquare \\\\n Tc = 220 K \\\\nThe mixture of I4/mmm YH4: \\\\na=2.708(1) Å, c=5.197(3) Å, V=38.1(1) \\\\nÅ3; \\\\nand Im-3m YH6 phase: \\\\na=3.542(2) Å, V=44.4(1) Å3. \\\\n5 (Q112) \\\\nY + H2. Pressurized to 185 (or 160 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser. \\\\n Fig.1 d \\\\n Tc ~ 214 K \\\\nThe mixture of I4/mmm YH4: \\\\na=2.766 Å, c=5.494 Å, V=42.0 Å3; \\\\nand Im-3m YH6 phase: \\\\na=3.570 Å, V=45.5 Å3; \\\\nand unidentified impurity(s). \\\\n6(Q142) \\\\nYD3 + D2. Pressurized to 194 (or 173 \\\\naccording to D2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nTc ~ 170 K \\\\n7 (Q121) \\\\nYD3 + D2. Pressurized to 198 (or 170 \\\\naccording to D2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nw/o electrodes \\\\nThe mixture of Im-3m YD6 phase: \\\\na=3.536(1) Å, V=44.2(1) Å3; \\\\nand unidentified impurity(s). \\\\nThe same sample on decompression \\\\ndown to 135 GPa (then the cell was \\\\nbroken). \\\\nThe mixture of Im-3m YD6 phase: \\\\na=3.585(6) Å, V=46.1(2) Å3 (at 135 \\\\nGPa); \\\\nand unidentified impurity(s). \\\\n8 (cell28) \\\\nY + H2. Pressurized to 23 GPa. Structural \\\\nchanges were followed by X-ray \\\\ndiffraction on compression up to 131 \\\\nGPa. w/o electrodes \\\\nFm-3m YH3 phase: \\\\na=4.928 Å, V=119.7 Å3 (at 23 GPa). \\\\nThe same sample after laser heating at \\\\n131 GPa and compression up to 140 GPa. \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.41 Å, V=85.8 Å3 (at 140 GPa). \\\\n9 (CT1) \\\\nY + H2. Pressurized to 85 GPa and heated \\\\nby pulsed laser. \\\\nw/o electrodes \\\\nFm-3m YH3 phase: \\\\na=4.536 Å, V=93.3 Å3. \\\\n10 (QL11) \\\\nY + H2. Pressurized to 105 GPa and \\\\nslightly heated by pulsed laser. \\\\nw/o electrodes \\\\nFm-3m YH3 phase: \\\\na=4.452(1) Å, V=88.2(1) Å3. \\\\nThe same sample heated by pulsed laser \\\\nseveral times at 105 and 130 GPa up to \\\\n\\\\uf07e1800 K. \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.388(1) Å, V=84.5(1) Å3 at 130 GPa. \\\\n11 (Q3) \\\\nYH3 + H2. Pressurized to 120 GPa and \\\\nslightly heated by pulsed laser. \\\\nw/o electrodes \\\\nMainly Fm-3m YH3 phase: \\\\na=4.407(1) Å, V=85.6(1) Å3; \\\\nand traces of unidentified impurity(s). \\\\n12 (W5) \\\\nYH3. Pressurized to 130 GPa and slightly \\\\nheated by pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.04 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YH3: \\\\na=4.373(6) Å, V=83.6(4) Å3; \\\\nand Fm-3m YH phase: \\\\na=3.986(2) Å, V=63.4(1) Å3. \\\\n13 (W7) \\\\nYD3. Pressurized to 135 GPa and slightly \\\\nheated by pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.02 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YD3: \\\\na=4.332(1) Å, V=81.3(1) Å3; \\\\nand Fm-3m YD phase: \\\\na=3.975(1) Å, V=62.8(1) Å3. \\\\n14 (Q1) \\\\nYH3 + H2. Pressurized to 150 GPa and \\\\nheated by pulsed laser up to \\\\uf07e1800 K. \\\\nw/o electrodes \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.284(3) Å, V=78.6(2) Å3. \\\\n15 (G2) \\\\nYD3. Pressurized to 170 GPa and slightly \\\\nheated by pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.5 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YD3: \\\\na=4.279(2) Å, V=78.3(2) Å3; \\\\nand Fm-3m YD phase: \\\\na=3.876(1) Å, V=58.2(1) Å3. \\\\n16 (W10) \\\\nYD3. Pressurized from 4 to 168 GPa at \\\\n\\\\uf07e100 K and then warmed. \\\\nw/o electrodes \\\\nFm-3m YD3 phase: \\\\na=4.260 Å, V=77.3 Å3; \\\\nand traces of Fm-3m YD phase. \\\\n17 (G1) \\\\nYH3. Pressurized to 170 GPa and heated \\\\nby pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.05 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YH3: \\\\na=4.286(2) Å, V=78.8(1) Å3; \\\\nand Fm-3m YH phase: \\\\na=3.901(1) Å, V=59.4(1) Å3. \\\\n18 (CM01) \\\\nY + H2. Pressurized to 215 (or 180 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser up to \\\\uf07e1800 K. \\\\nw/o electrodes \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.083 Å, V=68.1 Å3. \\\\n \\\\n', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce co"},"Similar Full Paper":{"kind":"string","value":"{'Local electronic properties of La3Ni2O7 under pressure': 'Title: Local electronic properties of La3Ni2O7 under pressure\\\\narXiv:2308.09044v1 [cond-mat.supr-con] 17 Aug 2023High-Tcsuperconductivity by mobilizing local spin singlets andpossible route to higher Tcin pressurized La3Ni2O7Qiong Qin1, 2 and Yi-feng Yang1, 2, 3, ∗1Beijing National Laboratory for Condensed Matter Physics and Institute of Physics,Chinese Academy of Sciences, Beijing 100190, China2University of Chinese Academy of Sciences, Beijing 100049, China3Songshan Lake Materials Laboratory, Dongguan, Guangdong 523808, China(Dated: August 21, 2023)We clarify the pairing mechanism of high-Tc superconductivity in bilayer La3Ni2O7 under highpressure by employing the static auxiliary field Monte Carlo approach to simulate a minimal effectivemodel that contains local dz2 inter-layer spin singlets and metallic dx2−y2 bands. Superconductivityis induced when the local spin singlet pairs are mobilized and attain long-distance phase coherenceby hybridization with the metallic bands. We find a dual role of hybridization that not only inducesglobal phase coherence but also competes with the spin singlet formation. This lead to a tentativephase diagram where Tc varies nonmonotonically with the hybridization, in good correspondencewith experimental observation. A roughly linear relation is obtained for realistic hopping andhybridization parameters: Tc ≈ 0.05J , where J is the inter-layer superexchange interaction. Weemphasize the peculiar tunability of the bilayer structure and propose that Tc may be furtherenhanced by applying uniaxial pressure along the c axis on superconducting La3Ni2O7. Our workprovides numerical evidences for the pairing mechanism of high-Tc superconductivity in La3Ni2O7and points out a potential route to achieve even higher Tc.Introduction.–The recently discovered high-Tc su-perconductivity in the bilayer nickelate La3Ni2O7 un-der high pressure [1–4] has stimulated intensive interestconcerning its basic electronic structures [5–11] and pos-sible pairing mechanism [12–26]. While first-principlesband calculations have predicted a Ni-d7.5 configurationwith almost fully-filled dz2 bonding band and two dx2−y2bands near quarter filling, it has also been argued thatthis weak coupling picture is not enough to explain thehigh Tc of about 80 K [23]. Indeed, strongly corre-lated electronic structure calculations have revealed well-formed dz2 moments with a large inter-layer superex-change interaction J via the O-pz orbital [10]. This laysthe basis for a strong coupling picture, where the dz2electrons provide preformed inter-layer spin singlets witha large pairing energy and the metallic dx2−y2 bands pro-vide a large phase stiffness. While it was suggested thata strong coupling of the two components could give riseto the high Tc [23, 27], other weak-coupling scenarioshave also been put forward to explain the pairing. It istherefore urgent to give more concrete calculations forqualitative or even quantitative comparison with experi-mental observations. In addition, one may be curious ifhigher Tc can be achieved in La3Ni2O7 by proper tuningbesides hydrostatic pressure.In this work, we propose that the high-Tc supercon-ductivity arises by mobilizing the local spin singlets ofdz2 electrons by hybridization with metallic dx2−y2 bands[23] and provide detailed numerical support for this pair-ing mechanism by performing static auxiliary field MonteCarlo simulations [28–34] on a minimal effective low-energy model. We construct a phase diagram showingqualitatively similar nonmonotonic evolution of Tc withincreasing hybridization strength as observed in experi-ment under pressure tuning. Our calculations reveal adual role of the hybridization in driving the supercon-ductivity. On the one hand, it helps to mobilize the localspin singlet pairs and induce a global phase coherencefor the superconductivity; on the other hand, it com-petes with the inter-layer superexchange interaction andtends to suppress the pairing strength. The overall goodconsistency with the experiments provides a strong sup-port of our scenario for the high-Tc superconductivity inLa3Ni2O7 under high pressure. We further find a roughlylinear relation for realistic hopping and hybridization pa-rameters, Tc ≈ 0.05J , and propose that higher Tc may beachieved by further applying uniaxial pressure along the caxis on superconducting La3Ni2O7. Our work points outthat mobilizing preformed spin singlets may be a generalroute for pursuing more high-Tc superconductors.Method.–We focus only on the pairing mechanismand study how superconductivity emerges based on thefollowing minimal effective bilayer Hamiltonian [23]:H = J∑iS1i · S2i −∑l〈ij〉σVij(d†liσcljσ + h.c)−∑l〈ij〉σ(tij + µδij)c†liσcljσ,(1)where dliσ (cliσ) is the annihilation operator of the dz2(dx2−y2) electron with spin σ on site i of layer l, andSli =12∑ss′ d†lisσss′dlis′ is the spin density operatorof dz2 electrons. The minimal model only includes theinter-layer antiferromagnetic superexchange interactionJ for dz2 electrons, the nearest-neighbor hopping pa-rameter tij and the chemical potential µ of the itiner-ant dx2−y2 electrons, and the in-plane nearest-neighbor2FIG. 1: Illustration of local dz2 inter-layer spin singlets (delectron) getting mobilized and attaining phase coherence bynearest-neighbor hybridization with metallic dx2−y2 bands (celectron). For clarity, the metallic orbitals are all shifted out-wards.hybridization Vij between two orbitals. Because the dz2and dx2−y2 wave functions are orthogonal on the sameNi-ion, the hybridization occurs mainly between nearest-neighbor sites and has opposite signs along x and y di-rections (Vi,i+x = −Vi,i+y = V ). An illustration of themodel is given in Fig. 1.Although simple, the above model covers all essen-tial ingredients for the superconductivity and gives aminimal description of dz2 spin singlet pairs mediatedby the inter-layer antiferromagnetic superexchange cou-pling. All other terms are not essential and dropped dueto strong onsite Coulomb repulsion [10, 23]. The absenceof direct hopping between dz2 orbitals indicates that theirspin singlets are local and cannot by themselves attainthe phase coherence to reach superconductivity. Super-conductivity can only emerge and become established byhybridization with the metallic dx2−y2 bands.To see how this mechanism is realized, we first decouplethe superexchange term via the Hubbard-Stratonovichtransformation [35]:JS1i · S2i →√2∆̄iψi +√2ψ̄i∆i +8∆̄i∆i3J, (2)where ψi =1√2(d1i↓d2i↑ − d1i↑d2i↓) represents the localinter-layer spin singlet of dz2 electrons at site i and ∆iis the corresponding auxiliary pairing field. However, di-rect Monte Carlo simulations typically suffer from severenegative sign problem. To avoid this issue, we adopt astatic approximation, ∆i(τ) → ∆i, by ignoring the tem-poral dependence of the pairing fields. The fermionicdegrees of freedom has a bilinear form and can be easilyintegrated out. Following Ref. [36], this gives an effec-tive action Seff(∆i) that depends solely on the complexpairing fields and can be simulated without the negativesign problem.The static auxiliary field Monte Carlo method has beenverified in previous studies of unconventional supercon-FIG. 2: (a) Intensity plot of the probabilistic distributionp(∆) of the local pairing fields on the complex plane ∆ =(∆x,∆y) for different temperatures T = 0.019, 0.027, 0.045 ata fixed hybridization V = 0.2. (b) Probabilistic distributionp(∆x) within a narrow cut |∆y | ≤ 0.01 for V = 0.03, 0.35,0.8 at a low temperature T = 0.001. (c) Evolution of thepeak position ∆max of p(∆x) at low temperature limit andthe onset temperature T∆ of local spin singlets as functionsof the hybridization V . T∆ marks the transition from the ringto a single maximum at the origin in the distribution plot (a).t is set to unity as the energy unit and J = 0.5.ductivity [28–30, 36–39]. It ignores dynamical fluctua-tions of the pairing fields but captures well their ther-mal and spatial fluctuations. The static approximationbreaks down at extremely low temperatures or near quan-tum phase transitions, but is suitable in our case to studyhow the phase coherence is established at finite temper-ature for the local dz2 spin singlets [36, 40]. We performthe Monte Carlo simulations on a 10×10 bilayer latticewith periodic boundary conditions, set t = 1 as the en-ergy unit, and choose J = 0.5t for the superexchangecoupling [9, 23]. For simplicity, the chemical potentialis fixed to µ = −1.3 so that the dx2−y2 and dz2 orbitalsare near quarter- and half-filled, respectively. The hy-bridization strength V is taken as the free parameter toconstruct the superconducting phase diagram.Local spin singlet pairs.–We first study the prob-abilistic distribution of the local spin singlets, p(∆i) =Z−1e−Seff(∆i), where Z is the partition function playingthe role of the normalization factor. A typical result isplotted in Fig. 2(a) on the complex plane (∆x,∆y) forthree different temperatures at V = 0.2. We find the dis-tribution clusters around the origin at high temperaturesbut gradually develops into a ring below a characteristictemperature T∆. The finite radius of the ring marks the3FIG. 3: (a) Intensity plot of the joint distribution betweenθ0 and θ(5,0) for V = 0.15 (upper panels) and 0.4 (lower pan-els) at three different temperatures, showing similar evolutionfrom uniform distribution to a stripe pattern. (b) The cor-responding phase mutual information I(5,0) as a function oftemperature for different hybridization parameters. The slopechange at low temperatures marks the crossover from short-to long-distance phase correlations and defines the tempera-ture scale T Ic . (c) Temperature dependence of dnv/dT for dif-ferent hybridization parameters. The maximum reflects thecharacteristic BKT transition for two-dimensional supercon-ductivity and defines the temperature scale T vc .formation of local dz2 spin singlet pairs. At low tem-peratures, its value reflects the intrinsic pairing strengthand may be estimated by plotting the distribution p(∆x)within a narrow cut |∆y| ≤ 0.01. This is plotted inFig. 2(b) for T = 0.001 and determined by the maxi-mum of the distribution. Interestingly, the peak positionmoves to a smaller ∆x with increasing hybridization V ,implying reduced pairing strength for strong hybridiza-tion. This is best seen in Fig. 2(c), where we plot T∆and ∆max as functions of V . Both quantities decreasemonotonically and reveal the competition between thelocal spin singlet formation and the hybridization.Phase coherence.–To see how superconductivityemerges from these local spin singlets, we study the long-distance phase correlations of the complex pairing fields∆i ≡ |∆i|eiθi . Figure 3(a) shows the joint distributionof the phase θi on two distant bonds, where θ0 is locatedon a chosen origin and θ(5,0) is on the bond at a distanceR = (5, 0). With lowering temperature, we see the evolu-tion from a uniform distribution at T = 0.045 to a stripefeature at T = 0.015. This indicates the gradual devel-opment of phase correlations between two distant bonds,a signature of global phase coherence between local spinsinglets. For comparison, we show the results for two hy-bridization parameters V = 0.15 and 0.4. They have verydifferent T∆ = 0.052 and 0.021, but the patterns of theirjoint distributions look quite similar for the same tem-perature. There is an obvious disparity between the spinsinglet formation and long-distance phase correlations.To clarify this, we quantify the phase correlations byintroducing the phase mutual information [41–46]:IR =∫dθ0dθR p(θ0, θR) lnp(θ0, θR)p(θ0)p(θR), (3)where p(θi) is the marginal distribution of the phase θiat site i and p(θ0, θR) is their joint probabilistic distribu-tion on two distant bonds at 0 and R after integratingout the pairing amplitude |∆i|. Figure 3(b) plots thephase mutual information I(5,0) as functions of the tem-perature for three different values of the hybridizationparameter V . In all cases, we find a gradual increase ofthe phase mutual information with lowering temperature.The increase grows rapidly in an intermediate tempera-ture range, marking a rapid development of phase cor-relations on two distant bonds. At a lower temperatureT Ic , a slope change is seen below which the phase mutualinformation grows less rapidly and seems to saturate to-wards some zero temperature limit. We will see thatT Ic may be identified as the superconducting transitiontemperature, at which the phase coherence is establishedbetween local spin singlet pairs on distant bonds.To further establish the superconducting transition, wealso calculate the vortex number [47], nv =∑i〈δwi,1〉,where the average is for all pairing configurations andwi is the winding number for θi → θi+x̂ → θi+x̂+ŷ →θi+ŷ → θi. We find nv increases rapidly in an inter-mediate temperature range. Its derivative dnv/dT isshown in Fig. 3(c) and defines another temperaturescale T vc at the maximum. Following the picture ofthe Berezinskii-Kosterlitz-Thouless (BKT) transition fortwo-dimensional superconductivity [48–50], the vortex-antivortex pairs are excited with temperature and be-come unbound across the transition, causing a rapid in-crease of nv around Tc. The peak in dnv/dT thereforemarks a characteristic feature of the BKT transition. In-triguingly, with increasing V , the peak position movesfirst towards higher temperatures (V = 0.23) but thenbackwards to lower temperatures (V = 0.5), indicating anonmonotonic variation of T vc in contrast to T∆.Superconducting phase diagram.–For better com-parison, we construct a superconducting phase diagramon the V -T plane in Fig. 4(a) and plot all three temper-ature scales, T∆, TIc , Tvc as functions of the hybridizationparameter V . Indeed, while T∆ decreases continuously4FIG. 4: (a) Theoretical phase diagram of the supercon-ductivity, showing all three temperature scales: T Ic from theslope change marking the long-distance phase coherence inthe phase mutual information plotted in Fig. 3(b), T vc fromthe maximum of dnv/dT marking the BKT transition in Fig.3(c), and T∆ from the probabilistic distribution of local pair-ing fields marking the transition from the ring distribution toa single maximum at the origin in Fig. 2(c). All results areobtained for J = 0.5. For comparison, the inset reproducesthe measured Tc in experiment under pressure on differentsamples [1, 3, 4], where the dashed line is a guide to the eye.(b) Variation of the superconducting transition temperatureTc estimated from Tvc as a function of the superexchange in-teraction J for different hybridization parameters V = 0.2,0.3, 0.5. t is taken as the energy unit.with increasing V , both T Ic and Tnc varies nonmonotoni-cally and collapse roughly on the same curve. The excel-lent coincidence between T Ic and Tvc provides further sup-port for the superconducting transition through globalphase coherence of local spin singlets and gives a consis-tent definition of Tc. We find a maximum Tc ≈ 0.025 atV ≈ 0.25. For smaller hybridization, Tc and T∆ behaveoppositely and there exists a wide intermediate temper-ature region Tc < T < T∆ where local spin singlet pairsexist but show no long-distance phase coherence. Thismarks a region of preformed pairs as previously proposedfor underdoped cuprates [51, 52]. We obtain the ratio2∆max/T∆ ≈ 4 − 6, a value close to those of pseudogapobserved in many experiments [53, 54]. Superconductiv-ity is only established when the local dz2 spin singlet pairsare get mobilized and attain phase coherence through hy-bridization with metallic dx2−y2 bands.For larger hybridization, the three temperature scalesbehave quantitatively similar, indicating that the super-conductivity is now constrained by the spin singlet for-mation rather than the phase coherence. The decrease ofTc with increasing V reflects the suppression of the pair-ing strength by the hybridization. The obtained ratio2∆max/Tc ≈ 7.5 − 9 is commonly observed in many un-conventional superconductors [55–57], and may be exam-ined in future experiment for superconducting La3Ni2O7.It should be noted that for two-dimensional supercon-ductivity, there is always a finite precursor region aboveTc. This is not plotted in our phase diagram but corre-sponds to the narrow region above T Ic in Fig. 3(b) wherethe phase mutual information grows rapidly with lower-ing temperature. In this regards, T∆ estimated from Fig.2(a) somewhat underestimates the onset temperature ofthe spin singlet pairs because of the large broadening ofthe ring. We will not go into more details on this be-cause we are mainly interested in the behavior of Tc andits comparison with experiment.Overall, our derived V -T phase diagram resemblesclosely those observed in experiment for La3Ni2O7 un-der pressure [1, 3, 4], where Tc exhibits a nonmonotonicvariation with pressure: it increases rapidly to near 80K from 10 GPa to 18 GPa and then decreases gradu-ally to about 50 K at 50 GPa as shown in the inset ofFig. 4(a). Our calculations suggest that this arises froma dual role of the hybridization, which mobilizes the lo-cal spin singlet pairs to induce global phase coherencebut at the same time competes to suppress their pairingstrength. It may also be illuminating to make some quan-titative estimate for direct comparisons. Taking t ≈ 0.5eV from first-principles calculations [5], our phase dia-gram for J/t = 0.5 yields a maximum Tc ≈ 0.025t, whichis roughly 140 K, the same order of magnitude as theexperimental Tc near 80 K considering that the real Tcmay be greatly reduced by many other factors beyond theminimal effective model. This overall agreement providesa strong support of our theory.Route to higher Tc.–Given the observed maximumTc near 80 K in experiment, it is desirable to ask if higherTc may be achieved upon proper tuning. Despite of somedelicacy in the pressure or hybridization tuning, some in-sight may still be gained by taking the liberty of modelcalculations. To explore other possibilities, we fix thehybridization parameter and change the superexchangeinteraction J . As shown in Fig. 4(b) for V = 0.2, Tc ex-hibits similar nonmonotonic behavior with increasing J .Evidently, the increase of Tc at small J is owing to the in-crease of the pairing strength, while its decrease at largeJ is constrained by the phase coherence due to hybridiza-tion. The maximum Tc can indeed be enhanced by tuningJ . For small J , Fig. 4(b) suggests a roughly linear rela-tion, Tc ≈ 0.05J , for realistic values of V and t. A crude5estimate for superconducting La3Ni2O7 yields J ≈ 0.5,which falls exactly in this region. Thus, a higher Tc maybe achieved by simply increasing the superexchange in-teraction J at fixed t and V . Fascinatingly, this mightactually be realized in experiment by further applyinguniaxial pressure along the c axis, since the hopping tand the hybridization V are both in-plane parameterswhile J is the inter-layer coupling. The fact that theymay be tuned separately highlights the importance ofthe bilayer structure of superconducting La3Ni2O7. Wesuggest future experiment to verify this simple but im-portant prediction of our minimal effective model.Conclusion.–We perform systematic calculations of aminimal effective model using the static auxiliary fieldMonte Carlo approach and propose that the high Tc su-perconductivity can be induced by mobilizing local spinsinglets through hybridization with metallic bands. Ourcalculations lead to the construction of a V -T phase di-agram where Tc varies nonmonotonically with increasinghybridization for fixed J . Our results agree well withexperiments and produce the correct order of magnitudefor the observed Tc. We further suggest the possibilityto achieve higher Tc by applying uniaxial pressure alongthe c axis on superconducting La3Ni2O7. Our theoryprovides a natural explanation of the high-Tc supercon-ductivity in La3Ni2O7 under high pressure, highlights theimportance of its bilayer structure, and points out a po-tential direction for future exploration of higher Tc.This work was supported by the Strategic PriorityResearch Program of the Chinese Academy of Sciences(Grant No. XDB33010100), the National Natural Sci-ence Foundation of China (Grants No. 11974397, No.12174429), and the National Key Research and Develop-ment Program of China (Grant No. 2022YFA1402203).∗ yifeng@iphy.ac.cn[1] H. Sun, M. Huo, X. Hu, J. Li, Z. Liu, Y. Han, L. Tang,Z. Mao, P. Yang, B. Wang, J. Cheng, D.-X. Yao, G.-M.Zhang, and M. Wang, Signatures of Superconductivitynear 80 K in a Nickelate under High Pressure, Nature(2023), https://doi.org/10.1038/s41586-023-06408-7.[2] Z. Liu, M. Huo, J. Li, Q. Li, Y. Liu, Y. Dai, X. Zhou,J. Hao, Y. Lu, M. Wang, and H.-H. Wen, ElectronicCorrelations and Energy Gap in the Bilayer NickelateLa3Ni2O7, arXiv:2307.02950.[3] J. Hou, P. T. Yang, Z. Y. Liu, J. Y. Li, P. F. Shan, L. Ma,G. Wang, N. N. Wang, H. Z. Guo, J. P. Sun, Y. Uwatoko,M. Wang, G.-M. Zhang, B. S. Wang, and J.-G. Cheng,Emergence of high-temperature superconducting phasein the pressurized La3Ni2O7 crystals, arXiv:2307.09865.[4] Y. Zhang, D. Su, Y. Huang, H. Sun, M. Huo, Z. Shan,K. Ye, Z. Yang, R. Li, M. Smidman, M. Wang, L.Jiao, and H. Yuan, High-Temperature Superconductiv-ity with Zero-Resistance and Strange Metal Behavior inLa3Ni2O7, arXiv:2307.14819.[5] Z. Luo, X. Hu, M. Wang, W. Wú, and D.-X. Yao, Bi-layer Two-Orbital Model of La3Ni2O7 under Pressure,arXiv:2305.15564.[6] Y. Zhang, L.-F. Lin, A. Moreo, and E. Dagotto, Elec-tronic Structure, Orbital-Selective Behavior, and Mag-netic Tendencies in the Bilayer Nickelate SuperconductorLa3Ni2O7 under Pressure, arXiv:2306.03231.[7] V. Christiansson, F. Petocchi, and P. Werner, Corre-lated Electronic Structure of La3Ni2O7 under Pressure,arXiv:2306.07931.[8] D. A. Shilenko and I. V. Leonov, Correlated Elec-tronic Structure, Orbital-Selective Behavior, and Mag-netic Correlations in Double-Layer La3Ni2O7 under Pres-sure, arXiv:2306.14814.[9] W. Wú, Z. Luo, D.-X. Yao, and M. Wang, Charge Trans-fer and Zhang-Rice Singlet Bands in the Nickelate Super-conductor La3Ni2O7 under Pressure, arXiv:2307.05662.[10] Y. Cao and Y.-F. Yang, Flat Bands Promoted byHund’s Rule Coupling in the Candidate Double-Layer High-Temperature Superconductor La3Ni2O7,arXiv:2307.06806.[11] X. Chen, P. Jiang, J. Li, Z. Zhong, and Y. Lu, Crit-ical Charge and Spin Instabilities in SuperconductingLa3Ni2O7, arXiv:2307.07154.[12] Q. G. Yang, H. Y. Liu, D. Wang, and Q. H. Wang,Possible s±-wave superconductivity in La3Ni2O7, arXiv:2306.03706.[13] F. Lechermann, J. Gondolf, S. Bötzel, and I. M. Eremin,Electronic Correlations and Superconducting Instabilityin La3Ni2O7 under High Pressure, arXiv:2306.05121.[14] H. Sakakibara, N. Kitamine, M. Ochi, and K. Kuroki,Possible High Tc Superconductivity in La3Ni2O7 underHigh Pressure through Manifestation of a Nearly-Half-Filled Bilayer Hubbard Model, arXiv:2306.06039.[15] Y. Gu, C. Le, Z. Yang, X. Wu, and J. Hu, Effective Modeland Pairing Tendency in Bilayer Ni-Based Superconduc-tor La3Ni2O7, arXiv:2306.07275.[16] Y. Shen, M. Qin, and G.-M. Zhang, Effective Bi-LayerModel Hamiltonian and Density-Matrix Renormaliza-tion Group Study for the High-Tc Superconductivity inLa3Ni2O7 under High Pressure, arXiv:2306.07837.[17] Y.-B. Liu, J.-W. Mei, F. Ye, W.-Q. Chen, and F.Yang, The s±-Wave Pairing and the Destructive Role ofApical-Oxygen Deficiencies in La3Ni2O7 Under Pressure,arXiv:2307.10144.[18] C. Lu, Z. Pan, F. Yang, and C. Wu, Interlayer Cou-pling Driven High-Temperature Superconductivity inLa3Ni2O7 under Pressure, arXiv:2307.14965.[19] Y. Zhang, L.-F. Lin, A. Moreo, T. A. Maier, andE. Dagotto, Structural Phase Transition, s±-WavePairing and Magnetic Stripe Order in the BilayeredNickelate Superconductor La3Ni2O7 under Pressure,arXiv:2307.15276.[20] H. Oh and Y.-H. Zhang, Type II t-J Model and SharedAntiferromagnetic Spin Coupling from Hund’s Rule inSuperconducting La3Ni2O7, arXiv:2307.15706.[21] Z. Liao, L. Chen, G. Duan, Y. Wang, C. Liu, R. Yu,and Q. Si, Electron Correlations and Superconductivityin La3Ni2O7 under Pressure Tuning, arXiv:2307.16697.[22] X.-Z. Qu, D.-W. Qu, J. Chen, C. Wu, F. Yang, W. Li,and G. Su, Bilayer t-J-J⊥ Model and Magnetically Me-diated Pairing in the Pressurized Nickelate La3Ni2O7,arXiv:2307.16873.[23] Y.-F. Yang, G.-M. Zhang, and F.-C. Zhang, MinimalEffective Model and Possible High-Tc Mechanism for6Superconductivity of La3Ni2O7 under High Pressure,arXiv:2308.01176.[24] K. Jiang, Z. Wang, and F.-C. Zhang, High TemperatureSuperconductivity in La3Ni2O7, arXiv:2308.06771.[25] Y. Zhang, L.-F. Lin, A. Moreo, T. A. Maier, and E.Dagotto, Trends of Electronic Structures and s±-WavePairing for the Rare-Earth Series in Bilayer Nickelate Su-perconductor R3Ni2O7,arXiv:2308.07386.[26] J. Huang, Z. D. Wang, and T. Zhou, Impurity and VortexStates in the Bilayer High-Temperature SuperconductorLa3Ni2O7, arXiv:2308.07651.[27] E. Berg, D. Orgad, and S. A. Kivelson, Route to High-Temperature Superconductivity in Composite Systems,Phys. Rev. B 78, 1 (2008).[28] M. Mayr, G. Alvarez, C. Şen, and E. Dagotto, PhaseFluctuations in Strongly Coupled d-Wave Superconduc-tors, Phys. Rev. Lett. 94, 217001 (2005).[29] Y. Dubi, Y. Meir, and Y. Avishai, Nature of theSuperconductor-Insulator Transition in Disordered Su-perconductors, Nature 449, 876 (2007).[30] M. Karmakar, Pauli Limited d-Wave Superconductors:Quantum Breached Pair Phase and Thermal Transitions,J. Phys. Condens. Matter 32, 405604 (2020).[31] K. Pasrija, P. B. Chakraborty, and S. Kumar, EffectiveHamiltonian Based Monte Carlo for the BCS to BECCrossover in the Attractive Hubbard Model, Phys. Rev.B 94, 165150 (2016).[32] J.-J. Dong, D. Huang, and Y.-F. Yang, Mutual Informa-tion, Quantum Phase Transition, and Phase Coherencein Kondo Systems, Phys. Rev. B 104, L081115 (2021).[33] A. Mukherjee, N. D. Patel, S. Dong, S. Johnston, A.Moreo, and E. Dagotto, Testing the Monte Carlo-MeanField Approximation in the One-Band Hubbard Model,Phys. Rev. B 90, 205113 (2014).[34] S. Liang, A. Moreo, and E. Dagotto, Nematic State ofPnictides Stabilized by Interplay between Spin, Orbital,and Lattice Degrees of Freedom, Phys. Rev. Lett. 111,047004 (2013).[35] P. Coleman, Introduction to Many-body Physics, (Cam-bridge University Press, Cambridge, England, 2015).[36] Q. Qin, J.-J. Dong, Y. Sheng, D. Huang, and Y.-F. Yang,Superconducting Fluctuations and Charge-4e PlaquetteState at Strong Coupling, Phys. Rev. B 108, 54506(2023).[37] Q. Han, T. Li, and Z. D. Wang, Pseudogap and Fermi-arcevolution in the phase-fluctuation scenario, Phys. Rev. B82, 052503 (2010).[38] Y. W. Zhong, T. Li, and Q. Han, Monte Carlo Study ofthermal fluctuations and Fermi-arc formation in d-wavesuperconductors, Phys. Rev. B 84, 024522 (2011).[39] D. K. Singh, S. Kadge, Y. Bang, and P. Majumdar,Fermi Arcs and Pseudogap Phase in a Minimal Micro-scopic Model of d-Wave Superconductivity, Phys. Rev. B105, 054501 (2022).[40] J.-J. Dong and Y.-F. Yang, Development of long-rangephase coherence on the Kondo lattice, Phys. Rev. B 106,L161114 (2022).[41] T. M. Cover and J. A. Thomas, Elements of InformationTheory (Wiley Series in Telecommunications and SignalProcessing) (Wiley-Interscience, USA, 2006).[42] A. Kraskov, H. Stögbauer, and P. Grassberger, Estimat-ing mutual information, Phys. Rev. E 69, 066138 (2004).[43] M. K. Varanasi, Estimation of the Information by anAdaptive Partitioning of the Observation Space Georges,IEEE Trans. Inf. Theory 45, 1315 (1999).[44] S. Khan, S. Bandyopadhyay, A. R. Ganguly, S. Saigal,D. J. Erickson, V. Protopopescu, and G. Ostrouchov,Relative performance of mutual information estimationmethods for quantifying the dependence among short andnoisy data, Phys. Rev. E 76, 026209 (2007).[45] M. I. Belghazi, A. Baratin, S. Rajeswar, S. Ozair, Y. Ben-gio, A. Courville, and R. D. Hjelm, Mutual InformationNeural Estimation, in Proceedings of the 35th Interna-tional Conference on Machine Learning (PMLR, 2018),Vol. 80, pp. 531-540.[46] B. Poole, S. Ozair, A. V. D. Oord, A. A. Alemi, andG. Tucker, On Variational Bounds of Mutual Informa-tion, in Proceedings of the 36th International Conferenceon Machine Learning, Long Beach, California (PMLR,2019), Vol. 97, p. 5171.[47] V. Drouin-Touchette, The Kosterlitz-Thouless PhaseTransition: An Introduction for the Intrepid Student,arXiv:2207.13748.[48] V. L. Berezinskii, Destruction of Long-Range Order inOne-Dimensional and Two-Dimensional Systems Pos-sessing a Continuous Symmetry Group. II. Quantum Sys-tems, Sov. Phys. JETP 34, 610 (1972).[49] J. M. Kosterlitz and D. J. Thouless, Ordering, Metasta-bility and Phase Transitions in Two-Dimensional Sys-tems, J. Phys. C Solid State Phys. 6, 1181 (1973).[50] J. M. Kosterlitz, The Critical Properties of the Two-Dimensional XY Model, J. Phys. C Solid State Phys.7, 1046 (1974).[51] B. Keimer, S. A. Kivelson, M. R. Norman, S. Uchida, andJ. Zaanen, From Quantum Matter to High-TemperatureSuperconductivity in Copper Oxides, Nature 518, 179(2015).[52] V. J. Emery and S. A. Kivelson, Importance of PhaseFluctuations in Superconductors with Small SuperfluidDensity, Nature 374, 434 (1995).[53] T. Yoshida, M. Hashimoto, S. Ideta, A. Fujimori, K.Tanaka, N. Mannella, Z. Hussain, Z. X. Shen, M. Kubota,K. Ono, S. Komiya, Y. Ando, H. Eisaki, and S. Uchida,Universal versus Material-Dependent Two-Gap Behav-iors of the High-Tc Cuprate Superconductors: Angle-Resolved Photoemission Study of La2−xSrxCuO4, Phys.Rev. Lett. 103, 037004 (2009).[54] C. Richter, H. Boschker, W. Dietsche, E. Fillis-Tsirakis,R. Jany, F. Loder, L. F. Kourkoutis, D. A. Muller, J.R. Kirtley, C. W. Schneider, and J. Mannhart, Inter-face Superconductor with Gap Behaviour like a High-Temperature Superconductor, Nature 502, 528 (2013).[55] H. Kim, Y. Choi, C. Lewandowski, A. Thomson, Y.Zhang, R. Polski, K. Watanabe, T. Taniguchi, J. Alicea,and S. Nadj-Perge, Evidence for Unconventional Super-conductivity in Twisted Trilayer Graphene, Nature 606,494 (2022).[56] G. Yao, M. C. Duan, N. Liu, Y. Wu, D. D. Guan, S.Wang, H. Zheng, Y. Y. Li, C. Liu, and J. F. Jia, Diamag-netic Response of Potassium-Adsorbed Multilayer FeSeFilm, Phys. Rev. Lett. 123, 257001 (2019).[57] D. S. Inosov, J. T. Park, A. Charnukha, Y. Li, A. V.Boris, B. Keimer, and V. Hinkov, Crossover from Weakto Strong Pairing in Unconventional Superconductors,Phys. Rev. B 83, 214520 (2011).', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on luteti"}}},{"rowIdx":80,"cells":{"Researcher Name":{"kind":"string","value":"Vitali Prakapenka"},"ORCID":{"kind":"string","value":"0000-0001-9270-2330"},"Researcher Query Keyword":{"kind":"string","value":"High-Pressure Yttrium Hydrides Synthesis and Characterization"},"Research Full Paper":{"kind":"string","value":"{'Superconductivity up to 243 K in the yttrium-hydrogen system under high pressure': 'Title: Superconductivity up to 243 K in the yttrium-hydrogen system under high pressure\\\\nSuperconductivity up to 243 K in yttrium hydrides under high pressure \\\\n \\\\nP. P. Kong1, V. S. Minkov1, M. A. Kuzovnikov1,5, S. P. Besedin1, A. P. Drozdov1, S. Mozaffari2, L. \\\\nBalicas2, F.F. Balakirev3, V. B. Prakapenka4, E. Greenberg4, D. A. Knyazev1 and M. I. Eremets1* \\\\n \\\\n1Max-Planck Institut für Chemie, Hahn-Meitner Weg 1, 55128 Mainz, Germany \\\\n2National High Magnetic Field Laboratory, Florida State University, Tallahassee, Florida 32310, USA \\\\n3NHMFL, Los Alamos National Laboratory, MS E536, Los Alamos, New Mexico 87545, USA \\\\n4Center for Advanced Radiation Sources, University of Chicago, 5640 South Ellis Avenue, Chicago, \\\\nIllinois, 60637, USA \\\\n5Institute of Solid State Physics Russian Academy of Sciences, 2 Academician Ossipyan str., \\\\nChernogolovka, Moscow District 142432, Russia \\\\n \\\\n \\\\nThe discovery of high-temperature conventional superconductivity in H3S with a critical temperature of \\\\nTc=203 K1 was followed by the recent record of Tc ~250 K in the face-centered cubic (fcc) lanthanum \\\\nhydride LaH102,3 compound. It was realized in a new class of hydrogen-dominated compounds having a \\\\nclathrate-like crystal structure in which hydrogen atoms form a 3D framework and surround a host atom of \\\\nrare earth elements4,5. Yttrium hydrides are predicted to have even higher Tc exceeding room temperature. \\\\nIn this paper, we synthesized and refined the crystal structure of new hydrides: YH4, YH6, and YH9 at \\\\npressures up to 237 GPa finding that YH4 crystalizes in the I4/mmm lattice, YH6 in Im-3m lattice and YH9 \\\\nin P63/mmc lattice in excellent agreement with the calculations5,6. The observed very high-temperature \\\\nsuperconductivity is comparable to that found in fcc-LaH102: the pressure dependence of Tc for YH9 also \\\\ndisplays a “dome like shape” with the highest Tc of 243 K at 201 GPa. We also observed a Tc of 227 K at \\\\n237 GPa for the YH6 phase. However, the measured Tcs are notably lower by \\\\uf07e30 K than predicted5,6. \\\\nEvidence for superconductivity includes the observation of zero electrical resistance, a decrease of Tc under \\\\nan external magnetic field and an isotope effect. The theoretically predicted fcc YH10 with the promising \\\\nhighest Tc>300 K was not stabilized in our experiments under pressures up to 237 GPa. \\\\n \\\\n \\\\n \\\\n \\\\n \\\\nIntroduction \\\\n One of the key characteristics of superconductivity is a critical temperature (Tc) below this \\\\ntemperature a metal can undergo an electronic transition towards a zero resistance state. Room \\\\ntemperature superconductors (RTSCs) have the potential to revolutionize our world. Perhaps the most \\\\nstraightforward way to reach RTSC can be found within the framework of conventional \\\\nsuperconductivity, where the theory is well established. In principle, Tc can be high in a metal displaying \\\\nsuitable parameters: an appropriate high-frequency phonon spectrum and a strong interaction between \\\\nelectrons and phonons at the Fermi level. The formula derived from the Bardeen-Cooper-Schrieffer and \\\\nMigdal-Eliashberg theories put no apparent limit in Tc 7. Tc, however, depends on a number of factors, \\\\nsuch as the details of the phonon and electronic spectra, which are difficult to estimate. Therefore, the \\\\ntheory provides only a general suggestion for the search of particular systems. Such systems should have \\\\nhigh characteristic phonon frequencies; for example, it can be realized with light atoms and strong \\\\ncovalent bonding. Based on these criteria, superconductivity was found in MgB28 but with a modest Tc of \\\\n39 K. The only material which has been always considered as a potential room-temperature \\\\nsuperconductor is metallic hydrogen9. Although superconductivity in metallic hydrogen has not yet been \\\\nexperimentally observed, this challenge has stimulated the idea to explore other hydrogen containing \\\\ncompounds such as CH4, SiH4, etc., as plausible candidates for the high Tcs.10 These compounds are \\\\ninsulators under ambient conditions, but they can be converted into covalently-bonded metals with the aid \\\\nof accessible pressures, that are much lower than those required for the metallization of pure hydrogen. \\\\nWhile high Tcs have not been found in the initially suggested hydrogen-dominant materials10, the \\\\napproach itself has proven to be very successful: superconductivity with Tc=203 K1 was found in \\\\nhydrogen sulfide, despite a modest hydrogen content in this compound. Further theoretical and \\\\nexperimental research in the H3S related family has not revealed any compound with a higher Tc. For \\\\ninstance, hydrogen selenide has a substantially lower Tc of \\\\uf07e105 K11. The discovery of new metal \\\\nsuperhydrides with the so-called clathrate-like structure of hydrogen atoms raised the calculated Tcs close \\\\nto the room temperature or even above it. In these hydrides, such as in the very first predicted CaH6 with a \\\\nTc of 235 K12, hydrogen atoms create a cage around the host calcium atom. Being connected to each other \\\\nin the network through weak covalent bonding, hydrogen atoms only weakly interact with the host metal \\\\natom through ionic bonding. The host atom supplies a charge to the hydrogen system, stabilizes the \\\\ncrystal structure and promotes metallization at a pressure below 200 GPa. The hydrogen network with \\\\nshort H-H distances in the clathrate-like structures is even closer to metallic atomic hydrogen in \\\\ncomparison to H3S, and can be considered as doped atomic hydrogen. The rapidly growing literature on \\\\nthis topic indicates that various transition metal, lanthanide or actinide elements are prone to form such \\\\nsuperhydrides and some of them exhibits superconductivity with much higher calculated critical \\\\ntemperatures4-6,12-15. The experimental evidence for the record Tc of \\\\uf07e250 K in the fcc-LaH10 at 150 GPa2,3 \\\\nhas confirmed the theoretical predictions and inspired experimentalists to synthesize new members of the \\\\nclathrate-like family of hydrides with promising high Tcs. In the present work, we studied the yttrium-\\\\nhydrogen system, which is the most attractive candidate for very high Tcs among all binary metal-\\\\nhydrogen systems theoretically studied so far. According to the calculations, superconducting fcc-YH10 \\\\nshould have extremely high Tc, that is as high as 303 K at 400 GPa5 or 305−326 K at 250 GPa13. In \\\\naddition to YH10, there are other phases predicted to display very high Tcs and to be stable at lower and \\\\nmore accessible pressures: hcp-YH9 with a Tc=253−276 K can be stabilized at P = 200 GPa5, and bcc-\\\\nYH6 with Tc=251−264 K that is stable already at 110 GPa6. In our experiments, we found the \\\\nsuperconducting phases YH6 and YH9 in agreement with the predicted structures5,6, but with Tcs \\\\nsignificantly lower than prediction by ~30 K. \\\\nResults and Discussion \\\\n Under ambient conditions, the yttrium hydride with highest hydrogen content is hcp-YH3. It is a \\\\nblack narrow-bandgap semiconductor with a metallic luster. When compressed in an inert medium, hcp-\\\\nYH3 converts to fcc-YH316. This pressure-induced phase transformation is extended under a wide pressure \\\\nrange of 10−25 GPa17,18. Further increase of pressure causes likely continuous metallization above \\\\n50 GPa19, as evidenced by the disappearance of the Raman spectrum and a significant drop in resistance. \\\\n Fcc-YH3 was predicted to be stable under high pressures up to 197 GPa20. We found that YH3 and \\\\nYD3 retain the fcc structure of the metallic lattice upon compression without medium up to P=180 GPa \\\\n(Samples 13, 15-18, SM Table 1) and do not exhibit the superconductive transition when subjected to \\\\npressures up to 170 GPa upon cooling down to 5 K. For these samples, we observed the appearance of a \\\\nnew fcc phase in addition to fcc-YH(D)3 under P =130–180 GPa. This new phase atom lattice volume is \\\\nsmaller, than that for fcc-YH(D)3 by ~5 Å3/Y, likely indicating reduced hydrogen content. A similar \\\\nphenomenon was reported earlier upon compression of a substoichiometric LaH3-x in an inert medium21, \\\\nwhere it was interpreted as resulting from a disproportionation reaction into hydrogen-rich stoichiometric \\\\ntrihydride and a hydrogen-poor solid solution. Given that our initial yttrium hydride samples were also \\\\nsubstoichiometric, the appearance of the hydrogen-depleted fcc phase in dry compression runs could also \\\\nresult from a disproportionation reaction. \\\\n When we compressed Y+H2 and Y+D2 mixtures, we observed the formation of YH3 and YD3 already \\\\nat 17 GPa (e.g. in sample 8), judging from a comparison between the atomic volume measured by XRD \\\\nand the literature data for YH318. In such experiments, we did not observe the hydrogen-depleted fcc \\\\nphase with a smaller unit cell parameter, which indicates a complete chemical reaction into the saturated \\\\nstoichiometric yttrium trihydride under excess hydrogen conditions. \\\\n The electrical properties and the structures of the yttrium hydrides with a H stoichiometry higher \\\\nthan YH(D)3 were of particular interest in the present study. Such hydrides were synthesized directly from \\\\nmixtures of Y (samples 3−5), YH3 (samples 1 and 2) or YD3 (samples 6 and 7) in excess of hydrogen \\\\n(deuterium) under high pressures. The chemical reaction occurs already at room temperature after several \\\\nweeks, but can be significantly accelerated with a pulsed laser that can heat the mixture up to \\\\uf07e2000 K. In \\\\nparticular, the I4/mmm YH4 and Im-3m YH6 were synthesized at a pressure of \\\\uf07e160 GPa with the aid of \\\\npulsed laser heating up to \\\\uf07e1000 K (samples 4 and 5 in SM Table 1). After several weeks under higher \\\\npressures between \\\\uf07e200 and 237 GPa, YH4 and YH6 can be synthesized already at room temperature \\\\n(samples 1 and 2 in SM Table 1). The P63/mmc YH9 phase can be synthesized starting from P\\\\uf07e184 GPa \\\\nbut only with the aid of pulsed laser heating (samples 1, 2 and 3, see details in SM Table 1). Higher \\\\npressures apparently promote phase transformation: YH9 is synthesized at 237 GPa even upon a subtle \\\\nsample heating kept below 700 K (no glowing observed) for sample 1 (see the structural details in Fig.2). \\\\n Samples prepared from YH3+H2 mixtures show much sharper transitions and a perfect zero \\\\nresistance state (see the details in Fig.1 (a)). For example, sample 1, which corresponds to the Im-3m \\\\nphase of YH6, was compressed to 237 GPa and kept at room temperature for 3 weeks, showed a sharp \\\\ntransition at 227 K to a zero resistance state (blue curve in Fig.1 (a)).For structural determination see \\\\nFigs.2 (d) and 2 (f). After keeping sample 2 under 201 GPa at room temperature for ~1 month, a Tc of 210 \\\\nK in the Im-3m YH6 phase was observed. With the aid of pulsed laser heating, the Im-3m YH6 phase was \\\\ntransformed into the P63/mmc YH9 phase with a Tc of 243 K (red curve in Fig. 1 (a), see Figs.2 (b) and 2 \\\\n(g) for the details concerning the identification of the structures). The superconductivity with a Tc of 243 \\\\nK could be ascribed to YH9, as samples 1 and 2 displayed Tc values around 210−220 K before the pulsed \\\\nlaser heating (or in the absence of the YH9 phase), and after the pulsed laser heating, the P63/mmc YH9 \\\\nphase was observed with Tc increasing to ~240 K. The main impurity phase in samples 1 and 2 before \\\\nlaser heating was the I4/mmm YH4 phase with c/a ≈ 1.9. This I4/mmm YH4 phase was found in many of \\\\nour samples (1−5), and its XRD pattern is shown in Fig.2 (a). Presently, we are unable to produce \\\\nreasonably pure I4/mmm YH4 to study its superconducting properties. According to calculations6, the \\\\nI4/mmm YH4 phase is superconducting with a Tc=84−95 K, which is considerably lower than those \\\\nmeasured in our samples and Tcs in the range 251−264 K, predicted for bcc-YH6. Thus, superconductivity \\\\nwith a Tc of 227 K and 210 K, as observed in samples 1 and 2 respectively, could be ascribed to the bcc-\\\\nYH6 phase. \\\\n The pressure dependence of Tc for the P63/mmc YH9 and the Im-3m YH6 phases from different \\\\nsamples is summarized in Figs.1 (b) and 1 (c). It is clearly seen in Fig.1 (b) that the pressure dependence \\\\nof Tc for YH9 has a “dome like shape” with the highest Tc at \\\\uf07e243 K under 201 GPa, which is similar to \\\\nthe value found for fcc-LaH102. The range of stability of the YH9 phase differs from the prediction5 – this \\\\nphase is stable at lower pressures. We found that the P63/mmc YH9 phase with higher Tc becomes stable \\\\nfrom at least 185 GPa. The unexpected abrupt drop in Tc(P) in the pressure range of 170−185 GPa, as \\\\nshown by the open black circles in Fig.1 (b), is probably related to the continuous distortion in the crystal \\\\nlattice as observed in SH3 at pressures below \\\\uf07e150 GPa1. In Fig.1 (c), for samples 1 and 2 (mixture of Im-\\\\n3m YH6 and I4/mmm YH4 phases), the onset Tcs were ascribed to the Im-3m YH6 phase. For sample 2, we \\\\ndefined the main superconducting transition at 210 K. A small drop in resistance was also observed at 220 \\\\nK, which is indicated by the smaller open black squares in Fig. 1 (c). However, sample 4 demonstrated a \\\\nhigher Tc~220 K with respect to sample 2. In addition to the Im-3m YH6 and the I4/mmm YH4 phases, \\\\nsample 4 also contained an unidentified complex phase (or a mixture of phases). Because the crystal \\\\nstructure and the stoichiometry of the impurities are not determined, it is not clear whether the \\\\nsuperconductivity observed in sample 4 is attributable to the YH6 phase. Recently, Troyan et al.22 \\\\nobserved superconductivity in the yttrium hydrides, synthesized through laser heating yttrium and \\\\nammonia borane under high pressures. Similarly to sample 4, their samples, revealed a Tc of 224 and 218 \\\\nK at 165 GPa, and comprised a complex mixture of phases; including the bcc-YH6, two new phases with \\\\nclaimed compositions YH7, Y2H15 and an unidentified impurity phase. On the basis of ab-initio \\\\ncalculations, Troyan et al. concluded that these phases should have lower Tcs, and assigned Tc ~220 K to \\\\nbcc-YH6. However, the poor agreement between the experimentally observed XRD patterns and the \\\\nproposed structural models (YH7 and Y2H15 phases) raises concerns about the reliability of their \\\\ninterpretation. Therefore, the superconducting properties of the pure YH6 phase at 160−200 GPa remain \\\\nopen. \\\\n Besides the observed drops in resistance to zero value upon cooling, superconductivity was verified \\\\nby the application of external magnetic fields up to \\\\uf06d0H = 9 T. Figure 1 (d) illustrates the dependence of \\\\nthe superconducting transition from sample 5 on an external magnetic field. Upper critical fields as a \\\\nfunction of temperature following the criterium of 90% of the resistance in the metallic state are shown in \\\\nthe inset of Fig. 1(d). The solid curve in the inset is an extrapolation to estimate the upper critical fields in \\\\nthe limit of zero-temperatures, after fitting the experimental data to the empirical relation: 𝐻𝑐2(𝑇) =\\\\n𝐻𝑐2(0) (1 − (\\\\n𝑇\\\\n𝑇𝑐\\\\n⁄ )\\\\n2\\\\n). This extrapolation yields Hc2(0) =110 T which is about 20 T larger than Hc2(0) \\\\nvalue in H3S 23. The zero resistance for the phases YH6 and YH9 (Fig.1 (a)) as well as the characteristic \\\\nshift of Tc as a function of the magnetic field (Fig.1 (d)) is a clear indication for superconductivity. \\\\n To determine the superconducting pairing mechanism, we substituted hydrogen with deuterium to \\\\nevaluate the effect on Tc. We observed a superconductivity with a Tc of 170 K in sample 6 which was \\\\nsynthesized through the pulsed laser heating of YD3 under deuterium atmosphere at temperatures >1000 \\\\nK and a pressure of 170 GPa. The structural determination still is in progress. \\\\n The crystallographic structure of all samples exhibiting superconducting transitions was determined \\\\nwith the aid of powder X-ray diffraction. The Rietveld refinement for the I4/mmm YH4, Im-3m YH6, and \\\\nP63/mmc YH9 crystal structures are shown in Figs. 2 (a), 2 (b), 2 (c), respectively. Figures 2 (d) and 2 (e) \\\\nare cake representations of the raw X-ray diffraction patterns collected for sample 1 before and after the \\\\npulsed laser heating. Figs. 2 (f) and 2 (g) demonstrate the changes in the powder X-ray diffraction \\\\npatterns during the heating of the mixture of YH3+H2 pressurized at \\\\uf07e200 and \\\\uf07e237 GPa for samples 2 \\\\nand 1, respectively. Pulsed laser heating initiates the phase transformation of YH4 into the YH6 phase, \\\\nand a subsequent transformation of YH6 into the YH9 phase at higher temperatures. \\\\n In order to estimate the stoichiometry of the newly synthesized yttrium hydrides, we studied YH3 in \\\\na wide range of pressures. The experimentally obtained compressibility for the crystal structure of fcc-\\\\nYH3/YD3 phase perfectly coincides with the theoretically calculated data14,20 (Fig.3 (a)). Taking into \\\\naccount the volume occupied by a Y atom in its pure metallic phase at the same pressure24, the additional \\\\nvolume caused by hydration, i.e. the difference between the volumes of YH3 and Y, is 6.7 Å3 at 150 GPa \\\\nand 5.1 Å3 at 215 GPa. Thus, the volume occupied by one hydrogen atom (VH) depends on pressure and \\\\nvaries from 2.2 Å3 to 1.7 Å3 for pressures ranging between 150 and 215 GPa. These estimated values for \\\\nVH are comparable to the one for the La-H system (1.9 Å3 at 150−178 GPa)2 and other metal-H systems25-\\\\n27. Using the calculated values of 1.6 Å3 for VH and 11.2 Å3 for the volume of yttrium derived from the \\\\nextrapolated data from the equation of state of metallic yttrium24, the stoichiometry calculated from the \\\\nexperimental diffraction data for the new yttrium hydrides are YH4.1, YH5.8, and YH8.8 at 237 GPa, \\\\nrespectively. The crystallographic structures of YH4, YH6, and YH9, agree perfectly with the theoretical \\\\npredictions5,6. The X-ray diffraction data for the volume of the crystal structure normalized with respect to \\\\nthe volume one Y atom as well as fragments of the crystal structures and coordination polyhedra for all \\\\nexperimentally found yttrium hydride phases are shown in Figure 3. The positions of the yttrium atoms \\\\nwere found directly from the diffraction data, whereas the hydrogen atoms were placed in the \\\\ngeometrically calculated positions based on the theoretical data5. \\\\n In spite of the very good agreement between the predictions and the experimental crystallographic \\\\nstructures, the measured Tcs for the Im-3m YH6 (Tc \\\\uf07e227 K) and the P63/mmc YH9 (Tc \\\\uf07e243 K) phases are \\\\nmarkedly lower than the predicted values of 251−264 K for YH66 and 253−276 K for YH95. Thus, only \\\\nthe fcc-YH10 phase can be expected to display RTSC with a predicted Tc \\\\uf07e 305−326 K13. However, we did \\\\nnot find YH10, in spite of extensive trials with high-pressure synthesis up to 237 GPa. Still there is a \\\\npossibility that this phase can be synthesized at higher pressures and temperatures. According to some \\\\npredictions, the fcc-YH10 phase is dynamically stable starting from 226 GPa15 or 220 GPa13. However, \\\\nother calculations5 suggest that the fcc-YH10 cannot be stabilized even at pressures as high as 400 GPa. \\\\nInstead, the hexagonal YH9 is energetically more favorable and lies on both convex hulls of formation \\\\nenthalpy and internal energy, while YH10 has a higher formation enthalpy and lies above the convex hull \\\\nby 24 meV/atom at 400 GPa that is associated to 1100 K5. The synthesis attempt of fcc-YH10 under \\\\nhigher pressures and temperatures is in progress. \\\\n \\\\nMethods \\\\n To synthesize the initial YH3 and YD3, yttrium metal of 99.9% purity was first annealed in a vacuum \\\\nof about 10-3 Torr at 400 °C for 4 h, and then exposed to hydrogen (or deuterium) gas at a pressure of \\\\nabout 100 bars at 200 °C for 24 h in a high-pressure Sievert-type apparatus. The sample treatment was \\\\ndone in an Ar glovebox to prevent oxidation. The reaction products were YH2.92(5) and YD2.87(5) as \\\\nindicated by their weight. We will further refer to these samples as YH3 and YD3 for brevity. The \\\\nsamples were analyzed through XRD with an Empyrean diffractometer at ambient conditions under a \\\\nKapton film. The lattice parameters of YH3 and YD3 were in reasonable agreement with the available \\\\ndata28. \\\\n In the diamond anvil cells (DACs), we typically synthesized yttrium hydride via a direct reaction \\\\nbetween yttrium (Alfa Aesar 99.9%) or YH3 and hydrogen (99.999%) at high pressures. For that, a piece \\\\nof Y or YH3 was placed into a hole drilled in an insulating gasket. The process of synthesis is the same as \\\\nthe one followed for lanthanum hydride2. The pressure, pulsed laser heating temperature, and the amount \\\\nof hydrogen surrounding the sample determined the composition of the yttrium hydrides. Superhydrides \\\\nwere synthesized only under an evident excess of hydrogen and high enough pressure. For the thermal \\\\ntreatment, one-sided pulsed radiation from a YAG laser was focused onto a spot having a diameter of 10 \\\\nµm. Some samples were prepared not from elemental yttrium as the starting material but from YH3 which \\\\nwas synthesized and analyzed as described above. One of the advantages of this method is to initially \\\\nhave a higher hydrogen content. To determine the isotope effect, we substituted hydrogen by deuterium \\\\n(99.75% purity). \\\\n Typically, the diamonds used in the DACs had a culet with a diameter of 20−35 µm and were \\\\nbeveled at 8° to a diameter of about 250 µm. A toroidal profile was machined at each culet by a focused \\\\nbeam of xenon ions (FERA3 TESCAN). Tantalum electrodes were sputtered onto the surface of one of \\\\nthe diamond anvils in a van der Pauw four-probe geometry and were covered with gold. A metallic gasket \\\\n(T301 stainless steel) was electrically separated from the electrodes by an insulating layer (a mixture of \\\\nepoxy and CaF2, MgO, CaSO4, cBN or Al2O3). The typical sample size was 5−10 µm. \\\\n We present resistance measurements upon warming the DACs as it yields a more accurate \\\\ntemperature reading: the cell was warmed up slowly (0.2 K min-1) under an isothermal environment (no \\\\ncoolant flow). The temperature was measured with an accuracy of about 0.1 K by a Si diode thermometer \\\\nattached to the DAC. All electrical transport measurements were performed with the electrical current set \\\\nin the range of 10-5-10-3 A. The pressure was measured through the H2 (D2) vibron scale29 if such a vibron \\\\ncan be observed, or otherwise from the diamond Raman edge scale30. The Tc was determined from the \\\\nonset of superconductivity – at the point of apparent deviation from the temperature dependence of the \\\\nresistance in the normal state metallic behavior. \\\\n We used three types of DACs. In particular, DACs with diameters of 20 mm and 8.8 mm were made \\\\nof nonmagnetic materials, suitable for measurements under magnetic fields using a 9 T Quantum Design \\\\nPhysical Property Measurement System (PPMS). The X-ray diffraction measurements were done with \\\\nwavelengths of 0.3344 Å and 0.2952 Å, an X-ray spot size ~3×3 µm, and Pilatus 1M CdTe detector at the \\\\nbeamline 13-IDD at GSECARS, Advanced Photon Source, Argonne National Laboratory (Chicago). \\\\nPrimary processing and integration of the powder patterns were made using the Dioptas software31. The \\\\nIndexing and refinement were done with GSAS and EXPGUI packages32. \\\\n \\\\nFigure captions \\\\nFigure 1. Superconducting transitions in yttrium hydrides. \\\\n(a) The temperature dependence of resistance for the Im-3m YH6 phase (blue curve, sample 1) and the \\\\nP63/mmc YH9 phase (red curve, sample 2). Samples 1 and 2 were synthesized from YH3 under hydrogen \\\\natmosphere. The configuration of the measurements is shown in the inset. After 3 weeks of maintaining \\\\nsample 1 at 237 GPa and at room temperature, a sharp transition appeared, indicating a superconducting \\\\ntransition with an onset of Tc= 227 K – blue curve. This transition corresponds to the Im-3m YH6 phase \\\\n(see the details of the structure in Fig.2 (d) and 2 (f)). After keeping sample 2 at 201 GPa and at room \\\\ntemperature for ~ 1 month, with the aid of pulsed laser heating, the P63/mmc YH9 phase was observed \\\\nwith a Tc =243 K (see Fig.2 (b) and 2 (g) for the structural characterization). \\\\n(b) The pressure dependence of Tc for the different samples belonging to the P63/mmc YH9 phase: \\\\nS1−filled magenta circles−sample 1 (after laser heating), S2 –filled black circles – sample 2 (after laser \\\\nheating), S3−filled red circles–sample 3. Details concerning the synthesis of the samples can be found in \\\\nthe SM Table 1. Data shown by open black circles were obtained by decreasing the pressure of sample 2. \\\\nThe open magenta circles correspond to the decompression of sample 1. The dotted line is a guide to the \\\\neyes. \\\\n(c) The pressure dependence of Tc for the different samples belonging to the Im-3m YH6 phase: S1−filled \\\\nmagenta squares−sample 1 (before laser heating), S2−filled black squares with main the superconducting \\\\ntransition at 210 K, smaller open black squares represent a small drop in resistance at 220 K−sample 2 \\\\n(before laser heating), S4−filled olive squares–sample 4. The details concerning sample synthesis can be \\\\nfound in SM Table 1. Samples 1 and 2 crystallize in the Im-3m YH6 and in the I4/mmm YH4 phases, with \\\\nthe Tc dominated by the Im-3m YH6 phase as indicated by the black polygon. Sample 4 crystallized in the \\\\nIm-3m YH6 and I4/mmm YH4 phases in addition to a complex unidentified phase (or a mixture of phases). \\\\n(d) Superconducting transition for sample 5 under an external magnetic field. This panel displays the \\\\nelectrical resistance as a function of the temperature under applied magnetic fields up to 9 T. For sample \\\\n5, R (T) was measured through a three terminal method, and the resulting background was subsequently \\\\nsubtracted. The temperature dependence of the observed upper critical fields (inset) was obtained from \\\\nthe data shown in Fig. 1 (d). An extrapolation to the lowest temperatures yields an \\\\uf07e110 T for the upper \\\\ncritical magnetic field in the limit of zero temperatures. \\\\nFigure 2. X-ray powder diffraction analysis of the synthesized yttrium hydrides through Rietveld \\\\nrefinement of the crystal structures. Black crosses are the experimental points, solid red lines are the \\\\nRietveld fits, and the blue solid lines represent the residuals. The blue, magenta and green ticks indicate \\\\nthe diffraction peaks corresponding to the tetragonal YH4, cubic YH6 and hexagonal YH9 phases, \\\\nrespectively. The used X-ray wavelengths (λ, Å) are indicated in the top right corners of each plot. The \\\\nrelative contribution between phases in a Rietveld refinement comprising two different crystalline phases \\\\nis summarized at the bottom right corner of each plot. Fit parameters for each refinement are shown under \\\\nthe blue residual graph. \\\\n(a) Rietveld refinement for the I4/mmm YH4 crystal structure found as a pure single-phase in some \\\\nregions of sample 4 subjected to 183 GPa. \\\\n(b) Rietveld refinement for the Im-3m YH6 crystal structure found in sample 2 under 201 GPa which was \\\\nkept at room temperature for 1 month. \\\\n(c) Rietveld refinement for the P63/mmc YH9 crystallographic structure found in sample 1 under 237 GPa \\\\nafter annealing for five cycles of very subtle pulsed laser heating (the temperature of each cycle was kept \\\\nbelow 700 K). \\\\n(d) Cake representation of the raw X-ray diffraction patterns collected for sample 1, which was kept at \\\\nroom temperature for 3 weeks. The three broad and highly spotted lines correspond to the Im-3m phase of \\\\nYH6 (this raw pattern corresponds to the black powder pattern in (f)). \\\\n(e) The same cake representation of sample 1 after five cycles of very subtle heating up to temperatures \\\\nbelow 700 K. The new narrow and dashed lines appearing after the heat treatment and correspond to the \\\\nP63/mmc phase of YH9 (this raw pattern corresponds to the green powder pattern in (f)). \\\\n(f) The changes in the powder patterns of sample 1 after successive cycles of very subtle heating (the \\\\ntemperature of each cycle is maintained below 700 K). The black powder pattern before any pulsed laser \\\\nheating corresponds to the nearly pure Im-3m phase of YH6. Each heating cycle results in the conversion \\\\nof YH6 into the higher YH9 hydride, as indicated by the appearance of new diffraction peaks related to the \\\\nP63/mmc lattice. Reflections from the Im-3m YH6 and P63/mmc YH9 phases are marked by c and h, \\\\nrespectively. \\\\n(g) Changes in the powder diffraction patterns after several cycles of pulsed laser heating for sample 2. \\\\nThe laser power and the corresponding temperature at each successive heating cycle was continuously \\\\nincreased up to 1750(15) K. At the beginning, the sample mainly consists of the Im-3m YH6 and I4/mmm \\\\nYH4 phases. Each heating cycle that maintains the temperature below ~1300 K initiates the \\\\ntransformation of the YH4 phase into YH6. Successive heating cycles to higher temperatures initiate the \\\\ntransformation of YH6 into the YH9 phase. \\\\n \\\\nFigure 3. Equation of state and crystal structures for the different YHx (0 ≤ x ≤ 9) phases. \\\\n(a) Unit cell volume normalized to the volume of one Y atom as a function of the pressure for the \\\\ndifferent yttrium hydrides as well as for pure yttrium. Data related to the same given phase are indicated \\\\nby the red polygons. Hexagons, stars, squares, triangles, and filled circles correspond to P63/mmc YH9, \\\\nIm-3m YH6, I4/mmm YH4, Fm-3m YH3, and distorted Fm-3m (hR24 and C2/m) yttrium, respectively. \\\\nRed, blue, green, magenta, purple, grey, navy, orange, azure, dark green, light grey, dark yellow, brown, \\\\ndark violet, black, light pink, light brown markers correspond to samples 1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, \\\\n13, 14, 15, 16, 17, and 18, respectively (see the SM Table 1). Filled black circles correspond to the \\\\nexperimental data for metallic yttrium24. Open black symbols correspond to the theoretically predicted \\\\ncrystallographic structures: open stars for YH65,6, open squares for YH46, and open triangles for YH314,20. \\\\n(b) Crystallographic structure for the different yttrium hydrides found in the experiment based on the X-\\\\nray diffraction data from the lattice of the heavier yttrium atoms. Hydrogen atoms were geometrically \\\\nplaced using their positions according to the theoretically predicted structures5. Yttrium and hydrogen \\\\natoms are coloured in blue and grey, respectively. The unit cell is indicated by red lines. The coordination \\\\npolyhedra (cages) and building polygons for these structures (YH3, YH4, YH6, YH9) are indicated by the \\\\nlight gray lines with the corresponding compositions shown in the row immediately below each structure. \\\\n \\\\n \\\\nAcknowledgements. \\\\nM.E. is thankful to the Max Planck community for the invaluable support, and U. Pöschl for the constant \\\\nencouragement. L.B. is supported by DOE−BES through award DE-SC0002613. S.M. acknowledges \\\\nsupport from the FSU Provost Postdoctoral Fellowship Program. The NHMFL acknowledges support from \\\\nthe U.S. NSF Cooperative Grant No. DMR−1644779, and the State of Florida. Portions of this work were \\\\nperformed at GeoSoilEnviro CARS (The University of Chicago, Sector 13), Advanced Photon Source \\\\n(APS), Argonne National Laboratory. GeoSoilEnviro CARS is supported by the National Science \\\\nFoundation−Earth Sciences (EAR−1634415) and Department of Energy−GeoSciences \\\\n(DE−FG02−94ER14466). This research used resources of the Advanced Photon Source, a U.S. Department \\\\nof Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne \\\\nNational Laboratory under Contract No. DE−AC02−06CH11357. \\\\n \\\\nCorrespondence and requests for materials should be addressed to M.E. \\\\n \\\\nReferences \\\\n1 Drozdov, A. P., Eremets, M. I., Troyan, I. A., Ksenofontov, V. & Shylin, S. I. Conventional \\\\nsuperconductivity at 203 kelvin at high pressures in the sulfur hydride system. Nature 525, 73 \\\\n(2015). \\\\n2 Drozdov, A. P. et al. Superconductivity at 250 K in lanthanum hydride under high pressures. \\\\nNature 569, 528 (2019). \\\\n3 Somayazulu, M. et al. Evidence for Superconductivity above 260 K in Lanthanum Superhydride \\\\nat Megabar Pressures. Phys Rev Lett 122, 027001 (2019). \\\\n4 Liu, H. et al. Dynamics and superconductivity in compressed lanthanum superhydride. Physical \\\\nReview B 98, 100102 (2018). \\\\n5 Peng, F. et al. Hydrogen Clathrate Structures in Rare Earth Hydrides at High Pressures: Possible \\\\nRoute to Room-Temperature Superconductivity. Phys Rev Lett 119, 107001 (2017). \\\\n6 Li, Y. et al. Pressure-stabilized superconductive yttrium hydrides. Sci Rep 5, 9948 (2015). \\\\n7 Allen, P. B. & Dynes, R. C. Transition temperature of strong-coupled superconductors \\\\nreanalyzed. Physical Review B 12, 905 (1975). \\\\n8 Nagamatsu, J., Nakagawa, N., Muranaka, T., Zenitani, Y. & Akimitsu, J. Superconductivity at 39 K \\\\nin magnesium diboride. Nature 410, 63 (2001). \\\\n9 Ashcroft, N. W. Metallic Hydrogen: A High-Temperature Superconductor? Physical Review \\\\nLetters 21, 1748 (1968). \\\\n10 Ashcroft, N. W. Hydrogen dominant metallic alloys: high temperature superconductors? Phys \\\\nRev Lett 92, 187002 (2004). \\\\n11 Mishra, A. et al. Novel Synthesis Route and Observation of Superconductivity in the Se-H System \\\\nat Extreme Conditions. APS March Meeting 2018 (2018). \\\\n12 Wang, H., Tse, J. S., Tanaka, K., Iitaka, T. & Ma, Y. Superconductive sodalite-like clathrate calcium \\\\nhydride at high pressures. Proc. Natl. Acad. Sci. 109, 6463 (2012). \\\\n13 Liu, H., Naumov, II, Hoffmann, R., Ashcroft, N. W. & Hemley, R. J. Potential high-Tc \\\\nsuperconducting lanthanum and yttrium hydrides at high pressure. Proc. Natl. Acad. Sci. 114, \\\\n6990 (2017). \\\\n14 Liu, L. L., Sun, H. J., Wang, C. Z. & Lu, W. C. High-pressure structures of yttrium hydrides. J Phys \\\\nCondens Matter 29, 325401 (2017). \\\\n15 Heil, C., di Cataldo, S., Bachelet, G. B. & Boeri, L. Superconductivity in sodalite-like yttrium \\\\nhydride clathrates. Physical Review B 99, 220502 (2019). \\\\n16 Palasyuk, T. & Tkacz, M. Hexagonal to cubic phase transition in YH3 under high pressure. Solid \\\\nState Communications 133, 477 (2005). \\\\n17 Machida, A., Ohmura, A., Watanuki, T., Aoki, K. & Takemura, K. Long-period stacking structures \\\\nin yttrium trihydride at high pressure. Physical Review B 76, 052101 (2007). \\\\n18 Machida, A. et al. X-ray diffraction investigation of the hexagonal–fcc structural transition in \\\\nyttrium trihydride under hydrostatic pressure. Solid State Communications 138, 436-440 (2006). \\\\n19 Nguyen, H., Chi, Z., Matsuoka, T., Kagayama, T. & K., S. Pressure-induced metallization of yttrium \\\\ntrihydride, YH3. J. Phys. Soc. Jpn. 81, SB041 (2012). \\\\n20 Li, Y. & Ma, Y. Crystal structures of YH3 under high pressure. Solid State Communications 151, \\\\n388 (2011). \\\\n21 Machida, A., Watanuki, T., Kawana, D. & Aoki, K. Phase separation of lanthanum hydride under \\\\nhigh pressure. Physical Review B 83, 054103 (2011). \\\\n22 Troyan, I. A. et al. Synthesis and Superconductivity of Yttrium Hexahydride Im-3m-YH6. arXiv e-\\\\nprints (2019). <https://ui.adsabs.harvard.edu/abs/2019arXiv190801534T>. \\\\n23 Mozaffari, S. et al. Superconducting phase diagram of H3S under high magnetic fields. Nature \\\\nCommunications 10, 2522 (2019). \\\\n24 Samudrala, G. K., Tsoi, G. M. & Vohra, Y. K. Structural phase transitions in yttrium under \\\\nultrahigh pressures. Journal of Physics: Condensed Matter 24, 362201 (2012). \\\\n25 Scheler, T. et al. High-pressure synthesis and characterization of iridium trihydride. Phys Rev Lett \\\\n111, 215503 (2013). \\\\n26 Pépin, C. M., Dewaele, A., Geneste, G. & Loubeyre, P. New Iron Hydrides under High Pressure. \\\\nPhys. Rev. Lett. 113, 265504 (2014). \\\\n27 Pépin, C. M., Geneste, G., Dewaele, A., Mezouar, M. & Loubeyre, P. Synthesis of FeH5: A layered \\\\nstructure with atomic hydrogen slabs. Science 357, 382 (2017). \\\\n28 Fedotov, V. K., Antonov, V. E., Bashkin, I. O., Hansen, T. & Natkaniec, I. Displacive ordering in the \\\\nhydrogen sublattice of yttrium trihydride. Journal of Physics: Condensed Matter 18, 1593 (2006). \\\\n29 Eremets, M. I. & Troyan, I. A. Conductive dense hydrogen. Nat Mater 10, 927 (2011). \\\\n30 Eremets, M. I. Megabar high-pressure cells for Raman measurements. Journal of Raman \\\\nSpectroscopy 34, 515 (2003). \\\\n31 Prescher, C. & Prakapenka, V. B. DIOPTAS: a program for reduction of two-dimensional X-ray \\\\ndiffraction data and data exploration. High Pressure Research 35, 223 (2015). \\\\n32 Toby, B. H. EXPGUI, a graphical user interface for GSAS. Journal of applied crystallography 34, \\\\n210 (2001). \\\\n \\\\nFigure 1. Superconducting transitions in yttrium hydrides. \\\\n(a) The temperature dependence of resistance for the Im-3m YH6 phase (blue curve, sample 1) and the \\\\nP63/mmc YH9 phase (red curve, sample 2). Samples 1 and 2 were synthesized from YH3 under hydrogen \\\\natmosphere. The configuration of the measurements is shown in the inset. After 3 weeks of maintaining \\\\nsample 1 at 237 GPa and at room temperature, a sharp transition appeared, indicating a \\\\nsuperconducting transition with an onset of Tc= 227 K – blue curve. This transition corresponds to the \\\\nIm-3m YH6 phase (see the details of the structure in Fig.2 (d) and 2 (f)). After keeping sample 2 at 201 \\\\nGPa and at room temperature for ~ 1 month, with the aid of pulsed laser heating, the P63/mmc YH9\\\\nphase was observed with a Tc =243 K (see Fig.2 (b) and 2 (g) for the structural characterization).\\\\n(b) The pressure dependence of Tc for the different samples belonging to the P63/mmc YH9 phase: \\\\nS1−filled magenta circles−sample 1 (after laser heating), S2 –filled black circles – sample 2 (after laser \\\\nheating), S3−filled red circles–sample 3. Details concerning the synthesis of the samples can be found in \\\\nthe SM Table 1. Data shown by open black circles were obtained by decreasing the pressure of sample 2. \\\\nThe open magenta circles correspond to the decompression of sample 1. The dotted line is a guide to \\\\nthe eyes. \\\\n(c) The pressure dependence of Tc for the different samples belonging to the Im-3m YH6 phase: S1−filled \\\\nmagenta squares−sample 1 (before laser heating), S2−filled black squares with main the \\\\nsuperconducting transition at 210 K, smaller open black squares represent a small drop in resistance at \\\\n220 K−sample 2 (before laser heating), S4−filled olive squares–sample 4. The details concerning sample \\\\nsynthesis can be found in SM Table 1. Samples 1 and 2 crystallize in the Im-3m YH6 and in the I4/mmm\\\\nYH4 phases, with the Tc dominated by the Im-3m YH6 phase as indicated by the black polygon. Sample 4 \\\\ncrystallized in the Im-3m YH6 and I4/mmm YH4 phases in addition to a complex unidentified phase (or a \\\\nmixture of phases). \\\\n(d) Superconducting transition for sample 5 under an external magnetic field. This panel displays the \\\\nelectrical resistance as a function of the temperature under applied magnetic fields up to 9 T. For sample \\\\n5, R (T) was measured through a three terminal method, and the resulting background was subsequently \\\\nsubtracted. The temperature dependence of the observed upper critical fields (inset) was obtained from \\\\nthe data shown in Fig. 1 (d). An extrapolation to the lowest temperatures yields an \\\\uf07e110 T for the upper \\\\ncritical magnetic field in the limit of zero temperatures. \\\\n150 180 210 240\\\\n220\\\\n230\\\\n240\\\\n250\\\\nPressure, GPa\\\\nYH\\\\n9\\\\nT\\\\nc\\\\n, \\\\nK\\\\n(b)\\\\nS1\\\\nS2\\\\nS3\\\\n0 50 100 150 200 250\\\\n0\\\\n20\\\\n40\\\\n60\\\\n80\\\\n100\\\\n120\\\\n150 200 250 300\\\\n0\\\\n20\\\\n40\\\\n60\\\\n80\\\\n\\\\uf06d\\\\n\\\\uf030\\\\nH\\\\n, \\\\nT\\\\nT , K\\\\nR\\\\ne\\\\ns\\\\nis\\\\nta\\\\nn\\\\nc\\\\ne\\\\n, \\\\n\\\\uf057\\\\nTemperature, K\\\\n 0 T\\\\n 1 T\\\\n 3 T\\\\n 5 T\\\\n 7 T\\\\n 9 T\\\\n(d)\\\\n150 200 250\\\\n200\\\\n220\\\\n240\\\\nPressure, GPa\\\\npure YH\\\\n6\\\\n+YH\\\\n4\\\\n phases\\\\nT\\\\nc\\\\n, \\\\nK\\\\n(c)\\\\nYH\\\\n6\\\\n+YH\\\\n4\\\\n+unidentified impurity\\\\nS2\\\\nS1\\\\nS4\\\\nS2\\\\n50 100 150 200 250 300\\\\n0.0\\\\n0.2\\\\n0.4\\\\n0.6\\\\n \\\\n \\\\nR*12\\\\nYH\\\\n6\\\\n T\\\\nc\\\\n = 227 K\\\\n P = 237 GPa\\\\nYH\\\\n9\\\\n T\\\\nc\\\\n = 243 K\\\\n P = 201 GPa\\\\nR\\\\ne\\\\ns\\\\nis\\\\nta\\\\nn\\\\nc\\\\ne\\\\n, \\\\n\\\\uf057\\\\nTemperature, K(a)\\\\nYH3 sample H2 electrode\\\\nFigure 2. X-ray powder diffraction analysis of the synthesized yttrium hydrides through Rietveld \\\\nrefinement of the crystal structures. Black crosses are the experimental points, solid red lines are the \\\\nRietveld fits, and the blue solid lines represent the residuals. The blue, magenta and green ticks \\\\nindicate the diffraction peaks corresponding to the tetragonal YH4, cubic YH6 and hexagonal YH9\\\\nphases, respectively. The used X-ray wavelengths (λ, Å) are indicated in the top right corners of each \\\\nplot. The relative contribution between phases in a Rietveld refinement comprising two different \\\\ncrystalline phases is summarized at the bottom right corner of each plot. Fit parameters for each \\\\nrefinement are shown under the blue residual graph. \\\\n(a) Rietveld refinement for the I4/mmm YH4 crystal structure found as a pure single-phase in some \\\\nregions of sample 4 subjected to 183 GPa. \\\\n(b) Rietveld refinement for the Im-3m YH6 crystal structure found in sample 2 under 201 GPa which was \\\\nkept at room temperature for 1 month.\\\\n(c) Rietveld refinement for the P63/mmc YH9 crystallographic structure found in sample 1 under 237 \\\\nGPa after annealing for five cycles of very subtle pulsed laser heating (the temperature of each cycle \\\\nwas kept below 700 K).\\\\n(d) Cake representation of the raw X-ray diffraction patterns collected for sample 1, which was kept at \\\\nroom temperature for 3 weeks. The three broad and highly spotted lines correspond to the Im-3m\\\\nphase of YH6 (this raw pattern corresponds to the black powder pattern in (f)).\\\\n(e) The same cake representation of sample 1 after five cycles of very subtle heating up to \\\\ntemperatures below 700 K. The new narrow and dashed lines appearing after the heat treatment and \\\\ncorrespond to the P63/mmc phase of YH9 (this raw pattern corresponds to the green powder pattern in \\\\n(f)).\\\\n(f) The changes in the powder patterns of sample 1 after successive cycles of very subtle heating (the \\\\ntemperature of each cycle is maintained below 700 K). The black powder pattern before any pulsed \\\\nlaser heating corresponds to the nearly pure Im-3m phase of YH6. Each heating cycle results in the \\\\nconversion of YH6 into the higher YH9 hydride, as indicated by the appearance of new diffraction peaks \\\\nrelated to the P63/mmc lattice. Reflections from the Im-3m YH6 and P63/mmc YH9 phases are marked \\\\nby c and h, respectively.\\\\n(g) Changes in the powder diffraction patterns after several cycles of pulsed laser heating for sample 2. \\\\nThe laser power and the corresponding temperature at each successive heating cycle was continuously \\\\nincreased up to 1750(15) K. At the beginning, the sample mainly consists of the Im-3m YH6 and \\\\nI4/mmm YH4 phases. Each heating cycle that maintains the temperature below ~1300 K initiates the \\\\ntransformation of the YH4 phase into YH6. Successive heating cycles to higher temperatures initiate the \\\\ntransformation of YH6 into the YH9 phase.\\\\n4 6 8 10 12 14 16 18 20\\\\n4 6 8 10 12 14 16 18 20\\\\n-1\\\\n0\\\\n1\\\\n2\\\\n3\\\\n4 6 8 10 12 14 16 18 20 22\\\\n4 6 8 10 12 14 16 18 20 22\\\\n-3\\\\n0\\\\n3\\\\n6\\\\n9\\\\n12\\\\n4 5 6 7 8 9 10 11 12 13 14\\\\n4 5 6 7 8 9 10 11 12 13 14\\\\n0\\\\n1\\\\n2\\\\n(c)\\\\n(b)\\\\nY 2d 0.333 0.667 0.750\\\\nY 2a 0 0 0\\\\na = 3.364(1) Å\\\\nc = 5.153(1)Å\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\n Iexp\\\\n Icalc\\\\n Iexp- Icalc\\\\n I4/mmm YH4\\\\nRp=19.0%, wR=37.9%\\\\nl=0.3344 Å\\\\na = 2.708(1) Å\\\\nc = 5.198(2) Å\\\\nY 2a 0 0 0\\\\n(a)\\\\na = 3.492(1) Å\\\\nl=0.2952 Å\\\\n Im-3m YH6\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\nRp=20.1%, wR=47.4%\\\\ntI:cI = 0.16 : 1\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\n P63/mmc YH9l=0.3344Å\\\\nRp=21.1%, wR=35.4%\\\\ncI:hP = 0.93 : 1\\\\n8 10 12 14\\\\n0\\\\n3\\\\n6\\\\n9\\\\n12\\\\n15\\\\n18 l=0.3344 Å\\\\n5\\\\n3\\\\n1\\\\nh\\\\nh\\\\nh h h\\\\nc\\\\nc\\\\nc\\\\nIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\nN\\\\no\\\\n. \\\\no\\\\nf \\\\nh\\\\ne\\\\na\\\\nti\\\\nn\\\\ng\\\\n c\\\\ny\\\\nc\\\\nle\\\\n0\\\\n(f)\\\\n7 8 9 10 11 12\\\\n0\\\\n1\\\\n2\\\\n3\\\\n4\\\\n5\\\\nt\\\\nl=0.2952 Å\\\\nh\\\\nh\\\\nh h\\\\nc\\\\nc\\\\nc\\\\nt t tIn\\\\nte\\\\nn\\\\ns\\\\nit\\\\ny\\\\n, \\\\na\\\\nrb\\\\n. \\\\nu\\\\nn\\\\nit\\\\ns\\\\n2q, °\\\\n\\\\uf07e300\\\\n\\\\uf07e650\\\\n\\\\uf07e1050\\\\n\\\\uf07e1300\\\\n\\\\uf07e1750\\\\nT, K\\\\n(g)\\\\n6 8 10 12 14\\\\n-180\\\\n-120\\\\n-60\\\\n0\\\\n60\\\\n120\\\\n180\\\\n6 8 10 12 14\\\\n-180\\\\n-120\\\\n-60\\\\n0\\\\n60\\\\n120\\\\n180\\\\n2q, °\\\\nA\\\\nz\\\\nim\\\\nu\\\\nth\\\\n, \\\\n°\\\\nA\\\\nz\\\\nim\\\\nu\\\\nth\\\\n, \\\\n°\\\\n2q, °\\\\n(d)\\\\n(e)\\\\nYH3\\\\nYH14\\\\n12 H4\\\\nYH4\\\\nYH18\\\\n4 H6 + 8 H4\\\\nYH6\\\\nYH24\\\\n6 H6 + 6 H4\\\\nYH9\\\\nYH29\\\\n6 H6 + 6 H5 + 6 H4\\\\n(b)\\\\n50 100 150 200 250\\\\n12\\\\n14\\\\n16\\\\n18\\\\n20\\\\n22\\\\n24\\\\n26\\\\n28\\\\nYH\\\\n4\\\\nYH\\\\n9\\\\nV\\\\no\\\\nlu\\\\nm\\\\ne\\\\n p\\\\ne\\\\nr \\\\na\\\\nto\\\\nm\\\\n, \\\\nÅ\\\\n3\\\\nPressure, GPa\\\\nY\\\\nYH\\\\n3\\\\nYH\\\\n6\\\\nhR24\\\\nC2/m\\\\n(a)\\\\nFigure 3. Equation of states and crystal structures of different YHx (0 ≤ x ≤ 9) phases.\\\\n(a) Unit cell volume normalized to the volume of one Y atom as a function of the pressure for the different \\\\nyttrium hydrides as well as for pure yttrium. Data related to the same given phase are indicated by the \\\\nred polygons. Hexagons, stars, squares, triangles, and filled circles correspond to P63/mmc YH9, Im-3m\\\\nYH6, I4/mmm YH4, Fm-3m YH3, and distorted Fm-3m (hR24 and C2/m) yttrium, respectively. Red, blue, \\\\ngreen, magenta, purple, grey, navy, orange, azure, dark green, light grey, dark yellow, brown, dark violet, \\\\nblack, light pink, light brown markers correspond to samples 1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, \\\\n17, and 18, respectively (see the SM Table 1). Filled black circles correspond to the experimental data for \\\\nmetallic yttrium24. Open black symbols correspond to the theoretically predicted crystallographic \\\\nstructures: open stars for YH6\\\\n5,6, open squares for YH4\\\\n6, and open triangles for YH3\\\\n14,20.\\\\n(b) Crystallographic structure for the different yttrium hydrides found in the experiment based on the X-\\\\nray diffraction data from the lattice of the heavier yttrium atoms. Hydrogen atoms were geometrically \\\\nplaced using their positions according to the theoretically predicted structures5. Yttrium and hydrogen \\\\natoms are coloured in blue and grey, respectively. The unit cell is indicated by red lines. The coordination \\\\npolyhedra (cages) and building polygons for these structures (YH3, YH4, YH6, YH9) are indicated by the light \\\\ngray lines with the corresponding compositions shown in the row immediately below each structure.\\\\nSample Synthesis conditions \\\\nElectrical \\\\nmeasurements \\\\nX-ray diffraction \\\\n1 (1010) \\\\nYH3 + H2. Pressurized to 255 (or 237 \\\\naccording to H2 vibron) GPa and kept for \\\\n3 weeks at room temperature. \\\\nTc = 227 K \\\\nFig.1c, magenta \\\\nsquare \\\\nIm-3m YH6, dominant phase: \\\\na=3.457(1) Å, V=41.3(1) Å3; \\\\nI4/mmm YH4: \\\\na=2.616 Å, c=5.184 Å, V=35.5 Å3. \\\\nThe same sample after 5 subtle laser \\\\nheating cycles (the temperature of each \\\\ncycle is below 700 K, no visible glowing). \\\\nTc = 237 K \\\\nFig.1b, filled \\\\nmagenta circle \\\\nP63/mmc YH9 phase: \\\\na=3.364(1) Å, c=5.153(1) Å, V=50.5(1) \\\\nÅ3; \\\\nwith remains of Im-3m YH6 and \\\\nI4/mmm YH4. \\\\n2 (Play) \\\\nYH3 + H2. Pressurized to 238 (or 201 \\\\naccording to H2 vibron) GPa and kept for \\\\n1 months at room temperature. \\\\nFig.1a, blue \\\\ncurve, Fig.1c, \\\\nblack square \\\\n Tc = 210 K \\\\nIm-3m YH6, dominant phase: \\\\na=3.492(1) Å, V=42.6(1) Å3; \\\\nwith I4/mmm YH4: \\\\na=2.656(1) Å, c=5.190(8) Å, V=36.6(1) \\\\nÅ3. \\\\nThe same sample after tens of laser \\\\nheating cycles up to 2000(10) K. \\\\nFig.1a, red \\\\ncurve, Fig.1b, \\\\nfilled black circle \\\\nTc = 243 K \\\\nP63/mmc YH9 phase: \\\\na=3.406 Å, c=5.210 Å, V=52.3 Å3; \\\\nwith remains of Im-3m YH6 and traces \\\\nof I4/mmm YH4. \\\\n3 (H120) \\\\nY + H2. Pressurized to 235 (or 184 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nFig.1b, filled red \\\\ncircle \\\\n Tc = 239 K \\\\nI4/mmm YH4, dominant phase: \\\\na=2.666 Å, c=5.194 Å, V=36.9 Å3; \\\\nIm-3m YH6 phase: \\\\na=3.529 Å, V=43.9 Å3; \\\\nP63/mmc YH9 phase: \\\\na=3.432 Å, c=5.251 Å, V=53.6 Å3. \\\\n4 (YH1) \\\\nY + H2. Pressurized to 189 (or 183 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nFig.1c, olive \\\\nsquare \\\\n Tc = 220 K \\\\nThe mixture of I4/mmm YH4: \\\\na=2.708(1) Å, c=5.197(3) Å, V=38.1(1) \\\\nÅ3; \\\\nand Im-3m YH6 phase: \\\\na=3.542(2) Å, V=44.4(1) Å3. \\\\n5 (Q112) \\\\nY + H2. Pressurized to 185 (or 160 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser. \\\\n Fig.1 d \\\\n Tc ~ 214 K \\\\nThe mixture of I4/mmm YH4: \\\\na=2.766 Å, c=5.494 Å, V=42.0 Å3; \\\\nand Im-3m YH6 phase: \\\\na=3.570 Å, V=45.5 Å3; \\\\nand unidentified impurity(s). \\\\n6(Q142) \\\\nYD3 + D2. Pressurized to 194 (or 173 \\\\naccording to D2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nTc ~ 170 K \\\\n7 (Q121) \\\\nYD3 + D2. Pressurized to 198 (or 170 \\\\naccording to D2 vibron) GPa. Heated by \\\\npulsed laser. \\\\nw/o electrodes \\\\nThe mixture of Im-3m YD6 phase: \\\\na=3.536(1) Å, V=44.2(1) Å3; \\\\nand unidentified impurity(s). \\\\nThe same sample on decompression \\\\ndown to 135 GPa (then the cell was \\\\nbroken). \\\\nThe mixture of Im-3m YD6 phase: \\\\na=3.585(6) Å, V=46.1(2) Å3 (at 135 \\\\nGPa); \\\\nand unidentified impurity(s). \\\\n8 (cell28) \\\\nY + H2. Pressurized to 23 GPa. Structural \\\\nchanges were followed by X-ray \\\\ndiffraction on compression up to 131 \\\\nGPa. w/o electrodes \\\\nFm-3m YH3 phase: \\\\na=4.928 Å, V=119.7 Å3 (at 23 GPa). \\\\nThe same sample after laser heating at \\\\n131 GPa and compression up to 140 GPa. \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.41 Å, V=85.8 Å3 (at 140 GPa). \\\\n9 (CT1) \\\\nY + H2. Pressurized to 85 GPa and heated \\\\nby pulsed laser. \\\\nw/o electrodes \\\\nFm-3m YH3 phase: \\\\na=4.536 Å, V=93.3 Å3. \\\\n10 (QL11) \\\\nY + H2. Pressurized to 105 GPa and \\\\nslightly heated by pulsed laser. \\\\nw/o electrodes \\\\nFm-3m YH3 phase: \\\\na=4.452(1) Å, V=88.2(1) Å3. \\\\nThe same sample heated by pulsed laser \\\\nseveral times at 105 and 130 GPa up to \\\\n\\\\uf07e1800 K. \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.388(1) Å, V=84.5(1) Å3 at 130 GPa. \\\\n11 (Q3) \\\\nYH3 + H2. Pressurized to 120 GPa and \\\\nslightly heated by pulsed laser. \\\\nw/o electrodes \\\\nMainly Fm-3m YH3 phase: \\\\na=4.407(1) Å, V=85.6(1) Å3; \\\\nand traces of unidentified impurity(s). \\\\n12 (W5) \\\\nYH3. Pressurized to 130 GPa and slightly \\\\nheated by pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.04 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YH3: \\\\na=4.373(6) Å, V=83.6(4) Å3; \\\\nand Fm-3m YH phase: \\\\na=3.986(2) Å, V=63.4(1) Å3. \\\\n13 (W7) \\\\nYD3. Pressurized to 135 GPa and slightly \\\\nheated by pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.02 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YD3: \\\\na=4.332(1) Å, V=81.3(1) Å3; \\\\nand Fm-3m YD phase: \\\\na=3.975(1) Å, V=62.8(1) Å3. \\\\n14 (Q1) \\\\nYH3 + H2. Pressurized to 150 GPa and \\\\nheated by pulsed laser up to \\\\uf07e1800 K. \\\\nw/o electrodes \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.284(3) Å, V=78.6(2) Å3. \\\\n15 (G2) \\\\nYD3. Pressurized to 170 GPa and slightly \\\\nheated by pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.5 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YD3: \\\\na=4.279(2) Å, V=78.3(2) Å3; \\\\nand Fm-3m YD phase: \\\\na=3.876(1) Å, V=58.2(1) Å3. \\\\n16 (W10) \\\\nYD3. Pressurized from 4 to 168 GPa at \\\\n\\\\uf07e100 K and then warmed. \\\\nw/o electrodes \\\\nFm-3m YD3 phase: \\\\na=4.260 Å, V=77.3 Å3; \\\\nand traces of Fm-3m YD phase. \\\\n17 (G1) \\\\nYH3. Pressurized to 170 GPa and heated \\\\nby pulsed laser. \\\\nMetallic \\\\nbehavior on \\\\ncooling with a \\\\nresidual \\\\nresistance of \\\\n\\\\uf07e0.05 Ohm at 5 \\\\nK. \\\\nThe mixture of Fm-3m YH3: \\\\na=4.286(2) Å, V=78.8(1) Å3; \\\\nand Fm-3m YH phase: \\\\na=3.901(1) Å, V=59.4(1) Å3. \\\\n18 (CM01) \\\\nY + H2. Pressurized to 215 (or 180 \\\\naccording to H2 vibron) GPa. Heated by \\\\npulsed laser up to \\\\uf07e1800 K. \\\\nw/o electrodes \\\\nUnidentified phase(s); \\\\nand remains of Fm-3m YH3 phase: \\\\na=4.083 Å, V=68.1 Å3. \\\\n \\\\n', 'Synthesis and Superconductivity of Yttrium Hexahydride Im$\\\\\\\\bar3$m-YH$_6$': 'Title: Synthesis and Superconductivity of Yttrium Hexahydride Im$\\\\\\\\bar3$m-YH$_6$\\\\nAbstract \\\\nFollowing the discovery of high-temperature superconductivity in the La–H system, where for the recently \\\\ndiscovered fcc-LaH10 a record critical temperature TC = 250 K was achieved [Drozdov et al., Nature, 569, 528 \\\\n(2019); Somayazulu et al., Phys. Rev. Lett. 122, 027001 (2019)], we studied the formation of new chemical \\\\ncompounds in the barium-hydrogen system at pressures of 75 to 173 GPa. Using in situ generation of hydrogen \\\\nfrom NH3BH3, we synthesized previously unknown superhydride BaH12 with a pseudocubic (fcc) Ba sublattice, \\\\nwhich was observed in a wide range of pressures from ~75 to 173 GPa in four independent experiments. DFT \\\\ncalculations indicate a close agreement between the theoretical and experimental equations of state. In addition \\\\nto BaH12, we identified previously known P6/mmm-BaH2 and possibly BaH10 and BaH6 as impurities in the \\\\nsamples. Ab initio calculations show that newly discovered semimetallic BaH12 contains H2, H3– molecular units \\\\nand detached H12 chains. Barium dodecahydride is a unique molecular hydride with metallic conductivity which \\\\ndemonstrates a superconducting transition around 20 K at 140 GPa in agreement with calculations (19–32 K). \\\\nThe interpretation of the multiphase XRD data was possible thanks to the development of new Python scripts for \\\\npostprocessing the \\\\nresults of evolutionary searches. These scripts help quickly identify the theoretical structures \\\\nthat explain the experimental data in the best way, among thousands of candidates. \\\\n \\\\nKeywords: Barium polyhydrides, high pressure, superconductivity, diamond anvil cell \\\\n \\\\n2 \\\\n \\\\nIntroduction \\\\nIn recent years, the search for new hydride superconductors with TC close to room temperature, attracts great \\\\nattention of researchers in the field of high-pressure materials science. Variation of pressure opens perspectives \\\\nfor synthesis of novel functional materials with unexpected properties.1 For example, according to theoretical \\\\nmodels,2–5 compression of molecular hydrogen over 500 GPa should lead to the formation of an atomic metallic \\\\nmodification with TC near room temperature. Pressures of 420-480 GPa were achieved in experiments with \\\\ntoroidal diamond anvil cells;6 however, for conventional high-pressure cells with a four-electrode electric setup, \\\\npressures above 200 GPa remain challenging. \\\\nIn 2004, Ashcroft7 suggested an alternative method of searching for high-TC superconductors using other \\\\nelements, metals or nonmetals, to precompress the hydrogen atoms, which should lead to a dramatic decrease in \\\\nthe metallization pressure. At the same time, because H is the lightest element in the periodic table, these \\\\nhydrogen-dominated compounds may have a high Debye temperature and strong electron-phonon coupling. \\\\nAshcroft’s work triggered many related theories and experiments. \\\\nA decade later Ashcroft’s idea found its experimental proof. Extraordinarily high superconducting transition \\\\ntemperatures were demonstrated in compressed 𝐹𝑚3̅𝑚-H3S8–11 (203 K at 150 GPa), 𝐹𝑚3̅𝑚-LaH1012–14 (>250 K \\\\nat 175 GPa), 𝐼𝑚3̅𝑚 -YH615 and P63/mmc-YH916 (224 and 243 K, respectively), 𝐹𝑚3̅𝑚 -ThH1017 (161 K at \\\\n174 GPa), and P63/mmc-CeH9 (~100 K).18 Recently, several semiempirical criteria for search for new hydride \\\\nsuperconductors were proposed, for example, the clathrate (H-cage) hydrogen substructure19,20 and “lability belt” \\\\nof superconductivity.21 From the theoretical point of view, only several metal polyhydrides satisfy these criteria \\\\nand have calculated TC above or close to 200 K. Mostly these are hydrides of alkali earth (Mg, Ca, Sr, Ba, etc.) \\\\nand early rare earth (La, Y, and Sc) metals. X-ray diffraction (XRD) experiments on compounds with heavy atoms \\\\nare more convenient because of better signal-to-noise ratio than on those with light elements like Li, Be, and Mg. \\\\nAnother practical reason to study hydrides of heavy metals is that their stabilization requires lower pressure. The \\\\nminimum stabilization pressure determines the maximum sample size that can be loaded into a DAC, which, \\\\ntogether with the atomic number, determines the exposure time and contrast of the diffraction pattern. These \\\\nreasons make the XRD experiments with polyhydrides of light elements complicated. \\\\nThe neighbor of lanthanum, barium is a promising element for the superhydride synthesis. The calculated \\\\nmaximum TC is only about 30–38 K20,21 for tetragonal P4/mmm-BaH6 stable at 100 GPa, which has a hydrogen \\\\nsublattice consisting of H2 molecules and H– anions.20 Lower barium hydride, BaH2, well-known for its \\\\nextraordinarily anionic (H–) conductivity,22 exists in Pnma modification below 2.3 GPa, whereas above 2.3 GPa \\\\nit undergoes a transition to hexagonal Ni2In-type P63/mmc phase.23 Chen et al.24 proposed a YbZn2-type structure \\\\nof BaH2 with Imma space group on the basis of ab initio calculations. However, this compound has not yet been \\\\nobserved in experiment. At pressures above 41 GPa, BaH2 transforms to P6/mmm modification, which metallizes \\\\nat over 44 GPa, but its superconducting TC is close to zero.25 \\\\nNone of the previously proposed structures of stable barium polyhydrides have a clathrate hydrogen sublattice. \\\\nSo far, no relevant experiments at pressures above 50 GPa have been reported. Due to analogy of the Ba–H and \\\\nLa–H systems, a recalculation of the Ba–H system is required and the existence of BaHn, where n ≥ 10, is \\\\nexpected (for instance, BaH10 and BaH12). Closing the gap of previous studies, in this work we experimentally \\\\nand theoretically investigated the chemistry of the barium-hydrogen system at pressures from 75 to 173 GPa. \\\\nResults and \\\\nDiscussion \\\\nTo investigate the formation of new chemical compounds in the Ba–H system at high pressures, we loaded \\\\nfour high-pressure diamond anvil cells (DACs #B0–B3) with a sublimated ammonia borane NH3BH3 (AB), used \\\\nas a source of hydrogen and a pressure transmitting medium. A tungsten foil preindented to a thickness of about \\\\n20 μm was used as a gasket. Additional parameters of the high-pressure diamond anvil cells are given in \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n \\\\n8 \\\\n \\\\n \\\\n \\\\nPressure, GPa a, Å b, Å c, Å V, Å3 VDFT, Å3 \\\\n140 5.500(1) 5.481(6) 5.539(3) 41.75 41.27 \\\\n138 5.526(9) 5.567(4) 5.500(3) 42.31 41.45 \\\\n136 5.535(2) 5.576(2) 5.517(7) 42.58 41.68 \\\\n132 5.546(4) 5.595(6) 5.510(8) 42.75 42.07 \\\\n128 5.542(1) 5.590(4) 5.505(0) 42.64 42.50 \\\\n123 5.560(1) 5.606(3) 5.525(9) 43.06 43.05 \\\\n111 5.602(6) 5.638(8) 5.571(2) 44.00 44.50 \\\\n99 5.706(5) 5.665(6) 5.637(4) 45.57 46.21 \\\\n93 5.758(3) 5.693(7) 5.700(6) 46.72 47.12 \\\\n86 5.759(9) 5.812(0) 5.717(9) 47.85 48.30 \\\\n75 5.831(0) 5.891(8) 5.797(2) 49.72 50.37 \\\\nh\\\\n-B\\\\naH\\\\n1\\\\n2\\\\n 140 4.0798 4.0798 5.7151 41.19 40.97 \\\\n136 4.0998 4.0998 5.7151 41.59 41.32 \\\\n132 4.1050 4.1050 5.7351 41.85 41.71 \\\\n111 4.1250 4.1250 5.7751 42.55 43.99 \\\\n \\\\n3. Synthesis at 90 GPa and electronic properties of BaH12 \\\\nIn the experiment with DAC #B3, we investigated the possibility to synthesize BaH12 at pressures below \\\\n100 GPa. After the laser heating of Ba/AB at 1600 K, the pressure in the cell decreased from 90 to 84 GPa and \\\\nthe resulting compounds were investigated using the synchrotron X-ray radiation (λ = 0.62 Å). The observed \\\\ndiffraction pattern is generally similar to those in the previous experiments with DAC #B1, except the presence \\\\nof the impurity phase h-BaH~12, whose reflections may be indexed by hexagonal P63/mmc or P63mc space groups \\\\n(lattice parameters a = 3.955(7) Å, c = 7.650(7) Å, V = 51.84 Å3 at 78 GPa) and whose cell volume corresponds \\\\nto approximately the same H content as in pseudocubic BaH12. For the main set of reflections, slightly distorted \\\\ncubic BaH12 is the best solution (\\\\n \\\\n \\\\n \\\\nBaH12 is the first known metal hydride with such a high hydrogen composition that is stable at such low \\\\npressures (~75 GPa). We performed further investigation of its electronic structure and the charge state of \\\\nhydrogen and barium atoms. The electron localization function (ELF) analysis 35 (\\\\n \\\\n \\\\n \\\\nSpin-polarized calculations demonstrate that in the pressure range of 50–200 GPa all barium hydrides are \\\\nnonmagnetic. The low electronic density of states N(E) in semimetallic Cmc21-BaH12, which looks typical for \\\\none-dimensional …H–H–H… chains (\\\\n \\\\n \\\\n \\\\nAs in all metal hydrides and polyhydrides, metal atoms donate some electrons to the hydrogen atoms. These \\\\nelectrons occupy the antibonding orbitals in the H2 molecules and weaken the H-H bonds. If zero electrons are \\\\ntransferred, the H2 molecules will persist and they will not contribute to superconductivity. If 1 electron is \\\\ntransferred, the hydride H- ions will be formed, with no H-H bonds and, again, little or no contribution to \\\\nsuperconductivity. At intermediate electron doping levels (as it has been found 19, 21, the optimum is ~0.3 electrons) \\\\nweak H-H bonds (e.g., as in clathrate polyhydrides) are formed. \\\\nIn BaH12, each H atom accepts few electrons, on average 0.16 electrons. As a result, H2 and H3 groups are still \\\\npresent in the structure, and we have a rather low TС. We think that at high pressures, due to dissociation of \\\\nmolecular groups, BaH12 may have network of weak H-H bonds (rather than discrete H2 and H3-groups) and, as \\\\na result, a much higher TС. Increasing the pressure will also facilitate further metallization of BaH12 and \\\\nsymmetrization of the hydrogen sublattice, increasing N(EF). To estimate the possible improvement, we \\\\ncalculated the superconducting parameters of I4/mmm-BaH12 and 𝐹𝑚3̅𝑚-BaH12, isostructural to YB12, structures \\\\nwhich were considered as possible solutions at the first step of the XRD interpretation. The calculations show \\\\nthat filling of pseudogap in N(E) makes it possible to reach TC ~ 200 K with λ = 3-4.5 in these compounds (see \\\\nSupporting Figures S39-40). \\\\nConclusions \\\\nIn our study of the high-pressure chemistry of the Ba–H system, we successfully synthesized in four \\\\nindependent DACs novel barium superhydride BaH12 with a pseudocubic crystal structure, stabilized in the \\\\npressure range of 75–173 GPa. The compound was obtained by the laser heating of metallic barium with an \\\\nexcess of ammonia borane compressed to 173, 160, 146, and 90 GPa. The Ba sublattice structure of BaH12 was \\\\nresolved using the X-ray synchrotron diffraction, evolutionary structure prediction, and several postprocessing \\\\nPython scripts, including an XRD matching algorithm. The discovered BaH12 has the highest hydrogen content \\\\n(>92 mol %) among all metal hydrides synthesized so far and unique metallic conductivity, localized in layers of \\\\nmolecular hydrogen. The experimentally established lower limit of stability of barium dodecahydride is 75 GPa. \\\\nThe third-order Birch–Murnaghan equation of state and unit cell parameters of BaH12 were found in the pressure \\\\nrange of 75–173 GPa: V100 = 45.47 ± 0.13 Å3, K100 = 305 ± 8.5 GPa, and 𝐾100\\\\n′ = 3.8 ± 0.48. The ab initio \\\\ncalculations confirm a small distortion of the ideal fcc-barium sublattice to Cmc21 (or even P21) space group, \\\\ndetermined by the presence of additional weak reflections in the diffraction patterns. The impurity phase analysis \\\\nindicates the possible presence of BaH6 and BaH10. According to the theoretical calculations and experimental \\\\nmeasurements, BaH12 exhibits semimetallic and superconducting properties with TC = 20 K at 140 GPa, and its \\\\ncrystal structure contains H2 and H3– groups. The \\\\nresults of these experiments confirm that the comparative \\\\nstability of superhydrides increases with the growth of the period number of a hydride-forming element in the \\\\nperiodic table.21 \\\\n12', 'Superconductivity up to 243 K in yttrium hydrides under high pressure': 'Title: Superconductivity up to 243 K in yttrium hydrides under high pressure\\\\nSuperconductivity up to 243 K in yttrium hydrides under high pressure \\\\n \\\\nP. P. Kong1, V. S. Minkov1, M. A. Kuzovnikov1,5, S. P. Besedin1, A. P. Drozdov1, S. Mozaffari2, L. \\\\nBalicas2, F.F. Balakirev3, V. B. Prakapenka4, E. Greenberg4, D. A. Knyazev1 and M. I. Eremets1* \\\\n \\\\n1Max-Planck Institut für C"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":81,"cells":{"Researcher Name":{"kind":"string","value":"Tabish Iqbal"},"ORCID":{"kind":"string","value":"0000-0002-2479-8729"},"Researcher Query Keyword":{"kind":"string","value":"Enhanced 1-Alkene Production via Chimeric UndB-Catalase Biocatalyst"},"Research Full Paper":{"kind":"string","value":"{'Functional production and biochemical investigation of an integral membrane enzyme for olefin biosynthesis': 'Title: Functional production and biochemical investigation of an integral membrane enzyme for olefin biosynthesis\\\\nAbstract ............................................................................................................................................ III Zusammenfassung ............................................................................................................................. V I. General \\\\nIntroduction .................................................................................................................. 1 1. A Few Words on Nomenclature ............................................................................................. 1 2. Iron(II)/α-Keto acid Dependent Enzymes .............................................................................2 3. Iron(III) and Iron(IV) Complexes ......................................................................................... 17 II. Motivation and Aim ................................................................................................................. 29 III. [FeIV(O)(Py5Me2H)]2+ as a TET Biomimetic.......................................................................... 31 1. \\\\nIntroduction and State of the Art ......................................................................................... 31 2. Motivation and Aim .............................................................................................................. 52 3. Synthesis of the Functional Model Complex C-6 ................................................................ 53 4. Stage 1: Nucleobases ............................................................................................................. 56 5. Stage 2: Nucleosides ............................................................................................................. 67 6. Stage 3: Dinucleotides .......................................................................................................... 74 7. Stage 4: Oligonucleotides .................................................................................................... 79 8. Synthetic Epigenetics and the Nucleobase Comparison Project........................................ 92 9. Miscellaneous: Syntheses and pH Dependence .................................................................. 101 IV. [FeIII(OH)(Py5Me2H)]2+ and Iron(IV)-oxido/Iron(II) Comproportionation ....................... 115 1. \\\\nIntroduction and State of the Art ........................................................................................ 115 2. Motivation and Aim ............................................................................................................. 117 3. Identification of [FeIII(OH)(Py5Me2H)]2+ as a Reaction Product ........................................ 118 4. Synthesis and Analysis of [FeIII(OH)(Py5Me2H)]2+ ............................................................. 124 5. Reactivity Studies with Organic Substrates ....................................................................... 128 6. Miscellaneous: Auto-Decomposition ................................................................................. 138 V. Immobilization, Py4PhMe2H, Py4OHMe2H, and Other Ligand Systems .............................. 143 1. \\\\nIntroduction and State of the Art ....................................................................................... 143 2. Motivation and Aim ............................................................................................................ 149 3. Immobilization of Py5Me2-X Ligands and Iron Complex Formation ................................. 151 4. The Py5(OR)2 Ligand System .............................................................................................. 165 5. The Py4OHMe2H Ligand and its Complexes ...................................................................... 176 6. Py4PhMe2H and Py3PhMe2H Ligands ................................................................................. 183 7. Q4Me2H – Combining Py5Me2H and TQA into One Ligand ............................................. 193 VI. Elucidating the Mechanism of Action of HPDL ................................................................ 195 1. \\\\nIntroduction and State of the Art ....................................................................................... 195 2. Motivation and Aim ........................................................................................................... 200 II 3. Metabolomics ...................................................................................................................... 201 4. Iron(IV)-oxido and Iron(III)-hydroxido Model Systems ................................................... 225 VII. PMS, PES, and Wurster’s Blue – A Short Mass Spectrometric Investigation .................... 231 1. \\\\nIntroduction and State of the Art ........................................................................................ 231 2. Motivation and Aim ............................................................................................................ 235 3. Phenazine Methosulfate and Phenazine Ethosulfate – PMS and PES .............................. 236 4. Wurster’s Blue – WB ........................................................................................................... 241 VIII. Summary and Outlook ....................................................................................................... 243 IX. Experimental Section ........................................................................................................... 251 1. Materials and \\\\nMethods ........................................................................................................ 251 2. General Procedures ............................................................................................................. 255 3. Synthetic Procedures – Ligands and Metal Complexes ..................................................... 258 4. Synthetic Procedures – Nucleobases ................................................................................... 312 5. Synthetic Procedures – Nucleosides and Dinucleotides ................................................... 327 6. Synthetic Procedures – HPDL Project ................................................................................ 338 7. Synthetic Procedures – Miscellaneous ............................................................................... 346 X. Appendix .................................................................................................................................. 351 1. List of Abbreviations ............................................................................................................ 351 2. Chapter III – Additional Data and Spectra ........................................................................ 356 3. Chapter IV – Additional Data and Spectra ........................................................................ 380 4. Chapter V – Additional Data and Spectra ...........................................................................381 5. Chapter VI – Additional Data and Spectra ........................................................................ 390 9. Chapter VII – Additional Data and Spectra ....................................................................... 417 10. NMR Spectra – Ligands and Complexes......................................................................... 418 11. NMR Spectra: Nucleobases ............................................................................................. 452 12. NMR Spectra: Nucleosides and Dinucleotides ............................................................. 469 13. NMR Spectra: HPDL Project ........................................................................................... 476 14. NMR Spectra – Miscellaneous ........................................................................................ 483 15. Crystallography .............................................................................................................. 486 XI.', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice par"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":82,"cells":{"Researcher Name":{"kind":"string","value":"Subhashini Murugan"},"ORCID":{"kind":"string","value":"0000-0002-4433-9191"},"Researcher Query Keyword":{"kind":"string","value":"Enhanced 1-Alkene Production via Chimeric UndB-Catalase Biocatalyst"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'The Catalytic Mechanism of Key Enzymes Involved in the Synthesis of Ergothioneine in Lentinula edodes': 'Title: The Catalytic Mechanism of Key Enzymes Involved in the Synthesis of Ergothioneine in Lentinula edodes\\\\nSEPTEMBER27th-30th 2022CROWNE PLAZABelgrade, Serbia~ ~ ~ELECTRONICABSTRACT BOOK~ ~ ~ ~ ORGANIZING COMMITTEE ~ ~ ~ ~ ~ ~ ~ ~ ~ SCIENTIFIC COMMITTEE ~ ~ ~ ~ ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Dear Colleagues, distinguished Guests, Ladies and Gentlemen!It is a great honour and pleasure to host you in Belgradefor the 11h International Medicinal Mushroom Conference. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Here in Belgrade, in the first IMMC in the third decade of its’ existence, we are proud to remind ourselves on the shared vision, strong belief and simultaneous e\\\\x1eorts of the three global pioneers in our scientific field, late Prof. Takashi Mizuno, Prof. Shu-Ting Chang and Prof. Soloman Wasser at the turn of a century, which resulted in what we can now proudly call the largest and highest quality international scientific gathering in the Medicinal Mushrooms domain and which we consider our joint heritage.In science, we receive new data and new results each day, constantly pushing our achievements towards the scientific community. This target remained unchanged for the whole, now more than two decades long history of our IMMC scientific gathering and all the activities of the International Society for medicinal mushrooms.Here, in Belgrade, we hope to be able to create a floor for scientists from di\\\\x1eerent continents, from di\\\\x1eerent countries, and from di\\\\x1eerent generations to meet, bring their knowledge and data on new scientific discoveries, and exchange on their methods and experience.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~I would like to share with you our most relevant impressions on the preparatory process of the IMMC11.It was everything but easy, however, the stakes were high an NO was not an option. As you all know we were in a position that we had to postpone this conference for a year. Still, IMMC11 Belgrade in September 2022, has been organized against many odds. From the beginning we struggled with the consequences of the COVID-19 global threat to the whole world, creating new future and changing our way of life. While being aware of very di\\\\x1eerent and sometimes rigorous quarantine measures, we could not previse extreme flight transportation costs increase, especially from China, all of you clearly aware of the immense importance of China, its’ researchers and businesspeople to our science and industrial production. We couldn’t predict the currently ongoing war in Europe, energy crisis and global protracted financial crisis which, among other, created strikes at airports. These were some of the reasons for the Organizing Committee to introduce hybrid conference for the IMMC11, with the intent to enable as many as possible colleagues and professionals to join and exploit benefits of this world class event in our scientific field.Despite described serious obstacles, we felt that we are to invest as much energy as needed to retain the continuity of the IMMC congress, our society heritage and high quality and spirit of our gatherings. That\\\\'s why we are immensely grateful to all of you who managed to join the congress and be here, in Belgrade, with us, to share your science, meet each other, exchange experience, and create newideas and projects.MESSAGE FROM THE ORGANIZERS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Message from the organizersWelcome! ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~During many previous congresses culture of poster presentations was not promoted enough, hence, invested energy and work of authors might not be visible enough, so our approach has been to put more emphasis on this important aspect, refresh it and made sort of a restart. We created and switched to electronic posters mode, visible throughout the conference. Six sessions with short oral communications in late afternoon hours are open for live presentation of authors, briefly presenting summary of their results for the audience and for discussions. (). Our expectations are that these sessions will create positive dynamics and intensify cooperation between the participants, and also contribute to future cooperation and projects.In these times of hardship, the least we can do, is to sincerely thank our colleagues and partners, representatives of companies from medicinal mushrooms science, production, and industry, for their benevolent and substantial support that enabled IMMC11 to be organized on truly appropriate level. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Here, on the IMMC11, besides scientific exchange, another important goal is to create opportunity for the broader international scientific community to converse in formal and informal settings, meet old friends and establish the new liaisons, as a legacy for the future, as well as to plan new scientific contacts and collaborations. The special emphasis on the IMMC11 is put on the young scientists – to be able to meet large number of renown international experts in our field. Organized in Europe after five years, this conference has true importance for Serbian, regional and European scientists. It is our hope that this Conference, organized in Belgrade end September 2022, will inject a new energy, and become a strong push for the Europe scientific community of medicinal mushrooms. Now the world center of medicinal mushrooms is here in Belgrade!We would like to use this opportunity to welcome all of you and to extend my best wishes for a successful and enjoyable stay in Belgrade and in Serbia.PROF. DR. MIOMIR NIKSIC IMMC11 Organizing Committee, President Institute of Food Technology and Biochemistry Faculty of Agriculture, University of Belgrade~ ~ ~ ~ ~ ~ ~ ~ ~ ~CONFERENCEPROGRAM• Prof. Ping Zhu / ChinaStudies on the biosynthesis of ergot alkaloidsProf. Zhi-Bin Lin / China• Dr. Siddhart Pramod Dubhashi / IndiaRole of Cordyceps in mild to moderate COVID infection01THE HISTORY OF ORGANIZING INTERNATIONAL MEDICINAL MUSHROOM CONFERENCES (IMMCS)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Prof. Miomir Nikšić1, Prof. Solomon P. Wasser2 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 Faculty of Agriculture University of Belgrade, Department of Industrial Microbiology, Serbia2 University of Haifa, Haifa , Israel~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~In science, each day we receive new data and new results, constantly pushing our achievements to-wards the scientific community. This target remained unchanged for the whole, now more than two decades long history of our IMMC scientific gathering and all the activities of the International Society for medicinal mushrooms.Shared vision, great belief and simultaneous efforts of the three global pioneers in our scientific field, late Prof. Takashi Mizuno, Prof. Shu-Ting Chang and Prof. Soloman Wasser at the turn of a century, resulted in what we can now proudly call the largest and highest quality international scientific gathering in the Medicinal Mushrooms domain and which we consider our joint heritage.Here, in Belgrade, we hope to be able to create a floor for scientists from different Continents, from different Countries, and from different generations to meet, bring their knowledge and data on new scientific discoveries, and exchange on their methods and experience.~ ~ ~ ~ ~ ~ ~ ~ ~ ~Before the first International Medicinal Mushroom Conferences (IMMC) were organized, the inaugu-ral issue of International Journal of Medicinal Mushroom (JMM) was published in January 1999 by Begell House. On the initiative and with the great effort of late Prof. Takashi Mizuno from Japan, who was the world\\\\'s guru and pioneer in the study of medicinal mushrooms, with Prof. Shu-Ting Chang, a world guru in the field of mushroom biotechnology and Prof. Soloman Wasser, famous fungal taxonomist, in September 2001, the 1st Conference was held on a high scientific and organizational levels in Kiev, Ukraine. The Chairman of this conference was Prof. Tetsuro Ikekawa from Japan, one of the developers of lentinan, together with Professor Giro Chihara, a developer of other medicinal mushroom products. At the Kiev Conference 350 scientists came together from 40 countries and it was decided to organize IMMCs every 2 years. Organizing International Medicinal Mushroom Conferences (IMMCs) was important to unite scientists in different fields of medicinal mushrooms from all over the world. 02THE HISTORY OF ORGANIZING INTERNATIONAL MEDICINAL MUSHROOM CONFERENCES (IMMCS)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The initiative sustained, the dream became reality and the new tradition has been created. The next Conferences followed, and, in July 2003 the 2nd IMMC was held in a beautiful place near Pathaya, Thailand under the Royal Family’s patronage. At the Thailand conference, approximately 1000 scientists participated, mainly from Asian countries. Dr Paul Stamets, founder, and President of Fungi Perfecti Co., than organized a very productive and successful 3rd IMMC in the State of Washington in the USA in October 2005, in beautiful Port Townsend, a historical city on the Pacific Coast but also a region with fantastic diversity of wild mushrooms. Approximately 300 scientists attended from 35 countries. Professor Franz Pohleven and Prof. Marin Berovic from Slovenia organized the 4th IMMC inLjubljana,in 2007. and this conference was very successful, too. In 2009 in Nontoong, China, under the patronage of Prof S.T. Chang and Prof. YU LI, and Head of China Chamber of Commerce of Foodstuffs and Native Produce, Mr. Zi-Qiang Liu, based on Alphay Bio-Techno-logical Company , organized the 5th IMMC Conference that was held on a very high scientific, organizational, and cultural levels, as well as gourmet delights. At Nantong IMMC 1200 participants were present. In 2011 in Zagreb, Croatia Dr Jakopovich President of Dr. Myko San Medicinal Mushroom Company together with his family, organized the 6th IMMC, a very productive, effective, and successful conference. In August 2013 in Beijing the group of Chinese Colleagues, again under the patronage of Professor S T. Chang, Prof LI, and Mr. Liu, organized a universally successful 7th IMMC medicinal mushroom conference. In Beijing, the Society received an invitation from colleagues based in South America, Dr. Carmenza Jaramillo from Colombia and Dr. Angela Amazonas from Brazil to hold the 8th IMMC in Manizales in 2015. The beautiful venue \"Termales el Otoño Hotel\" near Manizales, where very productive meeting has been organized, become event that will always be remembered.Professor Giuseppe Venturella organized the 9th Medicinal Mushroom conference in Palermo in 2017.We were delighted to attend in Sicily, Palermo near the volcano Etna, a very famous historical place.03THE HISTORY OF ORGANIZING INTERNATIONAL MEDICINAL MUSHROOM CONFERENCES (IMMCS)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Society accepted the invitation of Chinese colleagues to organize huge, 10th, jubilar huge International Medicinal Mushroom Conference in 2019 in China based on Alphay Biotechnological Company in Nantong. It was then, in 2019, in China, when we received an invitation to organize the next 11th IMMC conference in Belgrade, Serbia in 2021, however, to be postponed to this year due to the Covid-19 pandemics. Here, on the IMMC11 besides scientific exchange, another important goal is to create opportunity for the broader international scientific community to converse in formal and informal settings, meet old friends and establish the new ones, as a legacy for the future, as well as to plan new scientific contacts and collaborations. The special emphasis on the IMMC11 is put on the young scientists – to be able to meet large number of renown international experts in our field. Organized in Europe after five years, this conference has true importance for Serbian, regional and European scientists. It is our hope that this Conference, organized in Belgrade end September 2022, will inject a new energy and become a strong push for the Europe scientific community of medicinal mushrooms. Now the world center of medicinal mushrooms is here in Belgrade!Welcome! 04KEYNOTE 03RECENT ADVANCES IN THE RESEARCH OF MUSHROOMOF DIETARY FIBER AND BIOACTIVE POLYSACCHARIDES~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Peter C.K. Cheung~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~School of Life Sciences, Th Chinese University of Hong Kong, Hong Kong, China~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Mushroom cell wall is a rich source of non-digestible polysaccharides that are regarded as dietary fiber with multifunctional health benefits. The developmental stages of different mushroom species including spores, mycelia, fruit bodies and sclerotia provide various types of polysaccharides with different chemical structure and physical properties that have significant impact to their bioactivities. Traditional mushroom dietary fiber is well-known for its immunomodulatory and anticancer effects, hypoglycemic and hypocholesterolemic as well as antioxidant activities. Emerging evidence has revealed that the mechanisms by which mushroom dietary fiber can impart its specific health effects are mediated by gut microbiota and the prebiotic properties of mushroom polysaccharides. In particular, recent research findings have shown that the preferential utilization of mushroom beta-glucans by gut bacteria is associated with the newly identified polysaccharide utilization loci (PULs). Mushroom beta-glucans are shown to be a potential high molecular weight prebiotic with longer fermentation time in the colon and bifidogenic property. Current findings have indicated the unique role of mushroom beta-glucans to be used as a natural, safe and biocompatible nanomaterials to act as nanocarrier of drugs. Because of its specific recognition by immune cells such as macrophages via cell surface receptors, mushroom beta-glucans can targeted deliver the therapeutic agents that it carries to specific cell types. However, there remains several challenges for the applications of mushroom dietary fiber/polysaccharides in the functional food industry.Moreover, future research on mushroom polysaccharides needs to be focused on the structure-function relationship as well as clinical studies using structurally well-characterized samples.LECTURE 01KEYNOTE05MUSHROOMS, FOOD, OR MEDICINE? OR BOTH?~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Marina Sokovic1~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 Institute for Biological Research „Siniša Stanković“, National Institute of Republic of Serbia, University of Belgrade, Bulevar despota Stefana 142, 11040 Belgrade, Serbia~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Mushrooms are rich sources of bioactive compounds. The potential health benefits associated with mushroom intake are well-known. The study of the role of mushroom bioactive compounds to fight diseases, including cancer, is of major interest for our society. In the present study we have focused on the Portobello variety of Agaricus bisporus mushrooms. The aim is to find out whether agronomic intervention during mushroom cultivation may increase the content of certain bioactive compounds in the mushrooms and ultimately increase its anticancer activity.The present study evaluated the effects of vitamin D2, selenium and glucans on anticancer activity. For this purpose, several Portobello crops were grown on commercial phase II compost following standard cultivation practice to correlate the increase in vitamin D2 and the anticancer activity. For this purpose, ultraviolet radiation (UVA, UVB and UVC) was used to increase vitamin D2. UV-lamps were placed at 25 cm from the casing layer in the mushroom beds and treated with 6-18 J/cm2 for periods ranging from 15 to 45 min. UVA and UVB were applied during cultivation while UVC radiation was applied to freshly harvested mushrooms. Mushroom samples were extracted by saponification process and the levels of vitamin D2 were determined by HPLC. A significant increase in vitamin D2 was detected when mushrooms were irradiated with UVB but was not found in UVA-irradiated specimens. This increase in vitamin D2 was not followed by a significant improvement on the anti-cancer activity, indicating that the antitumor activity was not due to vitamin D2 contents.In addition, different crop trials were performed to evaluate the influence of selenium treatment during cultivation on the anti-cancer activity of cultivated Portobello. Sodium selenate was incorporated with irrigation water to raise the selenium contents of the mushrooms. Five separate treatments were per-formed through drench applications onto the casing layer prior to primordial formation, with water containing.LECTURE 02KEYNOTE06MUSHROOM MEDICINE: LATEST NEWS, SCIENCE, PRODUCT CONTROVERSIES, AND THERAPEUTIC USES~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Dr. Christopher Hobbs Ph.D.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~chrisrhobbs@gmail.com~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Author of the ground-breaking text, Medicinal Mushrooms (1989, 1995) and the recently released Medicinal Mushrooms, the Essential Guide (2021), Dr. Hobbs will discuss the historical and modern uses of fungi for medicine, and their place in modern integrative health care and medicinal practices. Review the complex cellular and molecular signaling pathways activated by exposure to cell wall constituents like beta-glucans, and the resulting immune changes that are the well-studied mechanisms of action enhancing immune vigilance against viruses and other pathogens.Other important constituent groups like di- and triterpenes, and phenolic compounds and their phar-macological effects will be discussed in light of their therapeutic actions and benefits.Mushrooms are among the highest fiber foods, and recent research on their health benefits as a source of prebiotics for helping to increase microbial diversity in the human microbiome, and associated health benefits will be reviewed. Practical considerations for making mushroom powder concentrates for addition to prepared foods like soups and smoothies will be detailed with samples to try.Finally, a special emphasis will be placed on how to collect or purchase fruiting bodies and make the best extracts and preparations, as well as choose the best commercial products including a discussion about recent controversies such as mycelium vs. fruiting body, and testing for starch content in mycle-lium products.LECTURE 03KEYNOTE07ENGINEERING ASPECTS IN CULTIVATION OF MEDICINALMUSHROOM BIOMASS IN SOLID STATEAND SUBMERGED BIOREACTORS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Prof.Dr. Marin Berović~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Faculty of Chemistry and Chemical Technology, University of Ljubljana, Slovenia~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Production of fungal fruit bodies using farming technology could hardly meet the demand of World market. Great interest for large scale production of various medicinal mushroom pharmaceutically active compounds requests the development of new comprehensive technologies. Research in physiology, basic and applied studies in fungal metabolism, process engineering and (pre)clinical studies in last two decades represent large contribution to the development of submerged and solid state cultivation of medicinal fungi biomass in bioreactors. In last three decades the fundamental aspects of solid state cultivation of various kinds of medicinal mushroom mycelia in various types of bioreactors was established. Solid state cultivation of various medicinal mushroom biomass is very close to fungal natural growth. Solid state cultivation in bio-reactors is well controlled comprehensive technology suitable for medium scale production especialy for recycling organic waste materials of different types. As the substrates various secondary wastes from wood, agriculture and fruit industry are successfully used. No fungal fruit boddies are produced. Final product represents delignolized, wooden material overgrown by medicinal fungi biomass enreach with proteins and various pharmaceutically products. Dryed and pulverized it could be directly used as a special veterinary remedies in a veterinary need.Development of comprehensive submerged technologies using stirred tank and air lift bioreactors are the most promising technologies for fast and large scale cultivation of medicinal fungi biomass and its pharmaceutically active products for human need. Pilot scale studies in this types of bioreactors represent the bridge and the balance between the gap of laboratory and industrial scale. In that sence it is not a surprise that most of pilot scale results and experiences remain uncovered industrial secrets. Comprehensive submerged cultivation represents fastest and the most suitable technology for a large scale production of medicinal mushroom biomass and its pharmaceutically active compounds for human use.This presentation is an overview of the engineering achievements in submerged and solid state cultivation in bioreactors.LECTURE 04KEYNOTE08MUSHROOMS AS A SOURCE OF BIO-BASED INGREDIENTS: FOOD AND COSMETIC APPLICATIONS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Lillian Barros~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 Centro de Investigação de Montanha (CIMO), Instituto Politécnico de Bragança, Campus de Santa Apolónia, 5300-253 Bragança, Portugal;2 Laboratório Associado para a Sustentabilidade e Tecnologia em Regiões de Montanha (SusTEC), Instituto Politécnico de Bragança, Campus de Santa Apolónia, 5300-253 Bragança, Portugal lillian@ipb.pt~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The growing world population and consumers’ awareness about what they eat and how it can affect their health, have boosted the interest in the consumption of mushrooms, and the investment of the food and pharmaceutical industries in the development of healthier products with bioactive assets. For centuries, mushrooms have been collected and consumed worldwide due to their valuable nutritional and medicinal properties and unique aroma, flavor, and texture. Their nutritional profile, mainly composed by carbohydrates, proteins, fibers, vitamins (B and C complexes), and minerals, makes them an excellent food product to be included in a daily diet.Besides, mushrooms are seen as functional ingredients and/or bases of nutraceuticals, given the presence of physiologically and biologically active substances, mostly in their fruiting bodies, mycelium and spores, such as polysaccharides, proteoglycans, terpenes, phenolic compounds, and others. These are responsible for several bioactive properties occurring in mushrooms, namely anti-proliferative, antioxidant, cholesterol reducing, and several others. Although mushroom extracts are widely studied for their bioactive value, little is known about their potential benefits in cosmetic products. Mushroom-derived metabolites can be sustainably used in the development of nutricosmetic and cosmeceutical formulations to suppress the severity of inflammatory skin diseases, to offer photoprotection to the skin, correct hyperpigmentation, among others. However, the use of mushroom extracts and their associated metabolites in these industries may be a challenge, since it includes several steps from extraction optimization, estimation of efficacy and safety, the use of micro and nano carriers, and the pros and cons associated with the use of extracts vs individual compounds. Also, and according with literature, mushrooms and their residues are a rich source of mycosterols, specially ergosterol (precursor of vitamin D2) and other steroids, bioactive molecules found in the nature that can exert different bioactive properties, being the only non-animal food source of vitamin D2, which is also formed during UV exposure, and whose deficiency can cause serious health problems. LECTURE 05KEYNOTE09MUSHROOMS AS A SOURCE OF BIO-BASED INGREDIENTS: FOOD AND COSMETIC APPLICATIONS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~For that, the fully investigation of these mushroom biomolecules and their incorporation into dairy, bakery and other types of food products represents a valuable opportunity to the food industry to provide functional foods to the world’s growing population. Therefore, the exploitation of mushrooms as a source of bioactive molecules has open new horizons making it possible to develop new bio-based food and cosmetic products. Aknowledgements: The authors are grateful to the Foundation for Science and Technology (FCT, Portugal) for financial support through national funds FCT/MCTES (PIDDAC) to CIMO (UIDB/00690/2020 and UIDP/00690/2020) and SusTEC (LA/P/0007/2021). L. Barros is also grateful to FCT, for her contract, through the institutional scientific employment program-contract. LECTURE 05KEYNOTE10ANTITUMOR EFFECT OF GANODERMA (LINGZHI)MEDIATED BY IMMUNOLOGICAL MECHANISMAND ITS CLINICAL APPLICATION~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Zhi-Bin Lin~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Departmant of Pharmacology, Peking University Health Science Center, Beijing, China~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The anti- tumor effect of Ganoderma(Lingzhi) is closely related to immunoregulation. Based on our research and other references this article discussed the antitumor effect of Ganoderma mediated by immunological mechanism including promoting the function of mononuclear- macro phages and natural killers promoting maturation and differentiation of dendritic cells increasing its antigen pre-sentation activating lymphocytes and increasing cytotoxicity of cytotoxin T lymphocyte promoting production of cytokines inhibiting tumor escape from immune surveillance. Also, clinical studies with immunological indexes were reviewed.Key words: Ganoderma; Lingzhi; polysaccharides; triterpens; immune; tumorLECTURE 06KEYNOTE11PSYCHEDELIC MUSHROOMS: RESEARCH, OPPORTUNITIES, AND THE FUTURE FOR PSILOCYBE IN MEDICINE~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Omoanghe S. Isikhuemhen1 and John C. Holliday ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 North Carolina A&T State University, Greensboro, NC 27411, USA.2 Ayla Bioscience Inc., 6646 Sierra Vista Lane, Carson City, NV 89702, USA.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The genus Psilocybe contains unique mushroom-forming species of fungi renowned for their hallucinogenic properties, exploited for neurotropic use, especially in many sacred religious ceremonies worldwide. There are about 300 Psilocybe species distributed worldwide in unique habitats like stems, leaves, seeds, earth, dung, sawdust, straw, dead wood, etc. Their hallucinogenic and psychoactive properties have earned them different common names like magic mushrooms, shrooms, goody-goody, psychedelics, etc. Psilocybe mushrooms are the principal source of naturally occurring psychedelics. The association of psychedelic use of Psilocybe with the hippie movement in the early 20th century and its classification as a Schedule 1 drug by the US government in 1970 brought scientific research on them to a stop and left lingering negative stigmatization until today. The compound psilocybin, commonly found in Psilocybe mushrooms and implicated in the psychedelic effects experienced when ingested, is biologically inactive. Upon ingestion, psilocybin is dephosphorylated to its active metabolite psilocin (4-hydroxy-N,N-dimethyltryptamine), which has mind-altering effects like euphoria, and visual and mental hallucinations, including changes in perception and perceived spiritual experiences. However, the prohibition on psychedelic drug research significantly delayed advances in medical knowledge on the therapeutic uses of agents such as psilocybin. A 2004 study on the use of psilocybin on advanced stage cancer patients gave results that reignited interest and significantly renewed efforts in psilocybin research and their exploration for psychedelic therapy. Results from clinical trials have shown significant reductions in depression and anxiety in cases of addiction, depression, and end-of-life mood disorders. Studies have also shown that psilocybin may reduce depression and anxiety associated with psychological crises due to a terminal diagnosis of advanced-stage cancer. Microdosing has become a term in everyday use. Users’ supposed positive experiences are the biggest driver of effort to change the negative popular opinion about psychedelic mushrooms and the laws prohibiting their use for research and therapy. Microdosing consists of three components: A) the use of a low dose below the perceptual threshold that does not impair ‘normal’ functioning of an individual. B) a procedure that includes multiple dosing sessions. C) The intention to improve well-being and enhance cognitive and/or emotional processes. LECTURE 07KEYNOTE12PSYCHEDELIC MUSHROOMS: RESEARCH, OPPORTUNITIES, AND THE FUTURE FOR PSILOCYBE IN MEDICINE~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Nevertheless, psilocybin cultivation, sale, and processing for medicinal use remain illegal worldwide, except in a few countries, including Jamaica and some South American countries. Countries in the west have started to decriminalize and legalize bits and pieces of cultivation, possession, recreational use, and medical research. Canada is taking the lead among Western countries on decriminalizing their use in medicine. In January 2022, Canada legalized psilocybin for prescription use, the first Western country to do so. However, magic mushrooms are prohibited in commerce in most of Europe and Asia. In general, laws and restrictions are still a significant obstacle for scientists seeking research grants to advance Psilocybe mushrooms’ research from cultivation to clinical applications. However, most countries where they are legal currently see enormous investments flowing from developed countries. Companies are trouping to these few countries to establish research from cultivation to clinical trials. They aim to perfect products and their use in the clinical treatment of various disorders to takefinancial advantage of the inevitable decriminalization and full-scale clinical use in western countries.Keywords: Hallucinogenic mushrooms, health benefits, psilocybe, psilocybin, psychedelicsLECTURE 07KEYNOTE13KEYNOTE 03CURRENT OVERVIEW OF PLEUROTUS SPECIES TAXONOMY, PHYLOGENETIC RELATIONSHIPS AND GLOBAL DISTRIBUTION~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Georgios I. Zervakis~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Agricultural University of Athens, Laboratory of General and Agricultural Microbiology, Iera Odos 75, 11855 Athens, Greece (zervakis@aua.gr)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The genus Pleurotus (Fr.) P. Kumm. includes taxa and species-complexes with a cosmopolitan distribution many of which are of high economic interest. In fact, commercial production of Pleurotus (“oyster”) mushrooms represents ca. 20% of the respective global production, and their cultivation is widespread throughout the world on a large variety of agricultural, agro-industrial and forestry residues. However, their study and subsequent exploitation was repeatedly hindered by problems related to ambiguous initial identifications (largely due to environmentally-influenced morphological characters) unclear and obsolete descriptions, erroneous use of taxonomic names, and conclusions based on fragmentary and/or not robust information. In addition, for many species, phylogenetic data are missing or are incomplete. Currently (June 2022), the number of Pleurotus taxa included in major da-tabases is several hundred, e.g., 769 in Index Fungorum and 830 in Mycobank; this is indicative of the perplexed status of taxonomy in the genus, and highlights the need to elucidate phylogenetic relation-ships among Pleurotus species, understand evolution issues and correctly assess the genus diversity.Although a substantial amount of work has been performed on certain groups, e.g., within the P. eryngii complex with members that are closely associated with plants of the family Apiaceae and are known to occur in the Old World only, or for coremia-forming Pleurotus (i.e., P. cystidiosus and allied taxa originating from various continents), species boundaries are not clear in several other groups. Such cases exist among the monomitic P. ostreatus, P. pulmonarius, P. floridanus, P. abieticola, P. eous, P. placentodes and P. populinus, or the dimitic P. cornucopiae, P. citrinopileatus and P. euosmus.Moreover, available data suggest that there is a large (yet untapped) diversity within tropical and subtropical species, e.g., the P. djamor complex (which includes several morphotaxa), P. giganteus sensu lato and P. tuber-regium. Last, limited phylogenetic information is available on species reported from the Neotropics, e.g., P. albidus, P. rickii and P. levis. In addition, this overview discusses indicative cases of uncertain/ambiguous taxonomic identity, and the erroneous use of names which are frequently met in scientific literature.LECTURE 01PLENARY14CONSERVATION, TAXONOMY, ECOLOGICAL DISTRIBUTION, AND POTENTIAL APPLICATION OF THE CULINARY-MEDICINAL MUSHROOM LECCINUM SCABRUM~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Giuseppe Venturella1,2, Maria Maddalena Cavalluzzi3, Giovanni Lentini3, Antonio Rosato3, Alexia Barbarossa3, Valeria Ferraro1, Fortunato Cirlincione1, Giulia Mirabile1, Ewa Zapora4, Marek Wolkowycki4, Marcin Stocki4, Pasqualina Colasuonno2, Maria Letizia Gargano2,5~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 Department of Agricultural, Food and Forest Sciences, University of Palermo, Palermo, Italy;2 Italian Society of Medicinal Mushrooms, Pisa, Italy;3 Department of Pharmacy - Pharmaceutical Sciences, University of Bari “Aldo Moro”, via E. Orabona, 4 70125 Bari (Italy); 4 Institute of Forest Sciences, Bialystok University of Technology, Poland;5 Department of Agricultural and Environmental Science, University of Bari “Aldo Moro”, Via Amendola 165/A, I-70126 Bari, Italy.E-mail: giuseppe.venturella@unipa.it~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Leccinum scabrum (Bull.) Gray, is a member of the family Boletaceae Chevall. It is an edible, mycorrhizal mushroom of potential application interest for both food and medicinal properties. Described in 1783 by French naturalist J.B.F. Bulliard under the name Boletus scaber, in 1821 it was included by S.F. Gray in the genus Leccinum. L. scabrum is mainly collected under birch trees, it prefers deciduous woods and is also found under Fagus sylvatica L. Fructification period extends from early summer to autumn, in grassy areas or with the presence of low bushes, in open spaces or at the edge of the woods. The cap is up to 15 cm width, when unripe, hemispheric then convex and finally flat. The surface of the cuticle is rather smooth, generally without depressions, fairly velvety, with a color ranging from off-white to light brown, to hazel, with yellowish tones, and is dry with dry weather, becoming somewhat viscous when it moistens. Sometimes, the cap have a fairly regular margin with small lighter or darker spots. Hymenium with white tubules then darker and gray-green in the ripe basidiomata. The pores are small with a rounded shape with a color, from whitish to gray and then brownish. Stipe 8-15 cm, 2-4 cm in diameter, firm and gradually fibrous, off-white, covered by dark small scales. Flesh off-white, unchanging when touched, cut or cooked. The flesh is tender but firm in the unripe specimens, while it tends to become soft when ripe. Pleasant light smell, taste sweetish. Basidiospores yellow-cinnamon, elongated shape, smooth and guttulate, 13-21 x 4-6 μm. L. scabrum is one of the most sought after and appreciated boletes. The characteristic of the meat that does not blacken, makes it particularly appreciated by many people. LECTURE 02PLENARY15CONSERVATION, TAXONOMY, ECOLOGICAL DISTRIBUTION, AND POTENTIAL APPLICATION OF THE CULINARY-MEDICINAL MUSHROOM LECCINUM SCABRUM~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~The firm consistency of the meat also contributes to its good edibility.Basidiomata of L. scabrum were collected within the Białowieża Forest (Poland), an ancient virgin forest (87,600 ha) with a unique biodiversity of fungi in Europe. Bialowieza Forest is the best pre-served forest ecosystem and the last low-land deciduous and mixed old-growth forest in Europe. Specimens of fungi of the genus Leccinum were identified on the basis of specialist literature using classical methods of taxonomic mycology. Evidence collections were made from the collected fruiting bodies and deposited at the Fungarium of the Institute of Forest Sciences (collection acronym in the Index Herbariorum - BLS).Results obtained from analyses performed on mushrooms powder obtained by drying of fresh basidiomata, show a relevant percentage of vitamins and minerals, among which vitamin D3, B2 and among minerals, sodium, potassium, iron and calcium. Remarkable is the content of carbohydrates, proteins and dietary fiber and the content of total sugars and total free amino acids. Polyunsaturated fatty acids represent the most marked value while saturated fatty acids represent the lowest value. Significant contents were found in vitamin D3 and B2 as well as sodium, potassium, iron, and calcium. Centesimal analysis shows a higher quantity of carbohydrates, proteins, dietary fiber and ashes compared to other foods. Furthermore, chemical composition of L. scabrum powder methanolic extract was also analyzed using gas chromatography with mass spectrometry. The main components of the extract were carbohydrates (79.61%), as well as fatty acids and their esters (6.59%). The extract also contained ergosterols (2.65%), polyalcohols (1.67%) and, amino acids (1.25%). In the extract was found presence of biologically active compounds belonging to hydroxy acids (e.g. malic acid, lactic acid, glyceric acid), dicarboxylic acids (e.g. succinic acid, fumaric acid, glutaric acid) and aromatic acids (i.e. benzeneacetic acid, 4-hydroxybenzeneacetic acid, benzoic acid).A preliminary study on its possible antibacterial effect was carried out by testing some extracts obtained under microwave irradiation against a panel of Staphylococcus aureus (ATCC and clinical iso-lates). Four different solvents with increasing polarity were used to extract the bioactive compounds of L. scabrum and the observed antibacterial activity was expressed as Minimal Inhibitory Concentration (MIC), assessed by microdilution method. The results obtained open the way for further investigation and for the activation of appropriate conservation strategies.Keywords: Biodiversity, Boletaceae, Taxonomy, Ecology, Conservation, Białowieza Forest, PolandLECTURE 02PLENARY16PROTEOMIC RESEARCH ON THE THERAPEUTIC PROPERTIES OF MEDICINAL MUSHROOMS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Boris Jakopović, PhD~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Dr Myco San – Zdravlje iz gljiva d.o.o. Croatia~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Medicinal mushrooms are increasingly being recognized as an important therapeutic modality in complementary oncology. Until now, more than 800 mushroom species have been known to possess significant pharmacological properties, of which antitumor and immunomodulatory properties have been the most researched. Besides a number of medicinal mushroom preparations being used as dietary supplements and nutraceuticals, several isolates from mushrooms have been used as official antitumor drugs in clinical settings for several decades. Proteomics is a large scale study of proteins, which is characterized by a hypothesis-free and comprehensive approach to studying novel mechanisms of potential therapeutics. Specifically, differential proteomics, also known as comparative or functional proteomics, studies the changes in proteome in different physiological or pathological statesbetween two or more samples. Cancer proteomics encompasses the identification and quantitative analysis of healthy tissue from neoplasia and can be used to identify markers for cancer diagnosis and treatment (biomarkers), monitoring disease progression, and identifying therapeutic targets serendipitously. Despite its complexity, proteomics is necessary for accurate characterization of pharmacological action. The complexity of cancer, which includes various pathways of its ontogeny and progression, tumor microenvironment, and therapeutic resistance mechanisms, poses demanding challenges that require a systems biology approach that is now increasingly beginning to re-examine the historical reductionist approach. In this regard, the previous research on cancer as well as on cancer therapeutics might be regarded as preliminary or partial and in need of a more in-depth study. However, various challenges remain. Besides the complexity and variability of biological material, i.e., medicinal mushrooms, the variability in the results may also be influenced by the model, either in vitro or in vivo, where the timing of the treatment (early vs. late models of disease; immunocompetent vs. nude mice xenografts), as well as the tumor model (orthotopic vs. heterotopic), can producedifferent results.The results are also highly dependent on the proteomic methods, which also differ in their sensitivity. However, these obstacles must be overcome in order to study complex diseases such as cancer.This review is focused on the current state of proteomic research into antitumor mechanisms of some of the most researched medicinal mushroom species, including Phellinus linteus, Ganoderma lucidum, Auricularia auricula, Agrocybe aegerita, Grifola frondosa, and Lentinus edodes, as whole body extracts or various isolates, as well as of complex extract mixtures.LECTURE 03PLENARY17FUTURE OF MUSHROOM FARMING~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ John Holliday~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Ayla Bioscence Inc, USA~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~The last 20 or 30 years has brought great change to the field of mushroom cultivation. Advances in equipment, chemistry, strains, and technology are driving ever increasing yields at lower input costs than ever before. By designing and building our farms using today’s knowledge, materials and techniques, mushroom farming is one of the most profitable agricultural endeavors known.To best understand these changes and look at how mushroom farms of today and tomorrow are evolving less as agricultural endeavors but more as Biological Manufacturing Processes, we need to first look at the past. How the mushroom industry first developed and has advanced up until now. The earliest recorded cultivation of mushrooms started around the year 800 AD, with growing wood ears (Auricularia) on natural logs in China. By the year 1200, shiitake was widely cultivated this way as well. First the bark was split lengthwise and peeled back from the wood. A hole was made with a hollow cylindrical tool attached to a handle, resulting in a tool that that looks like a hollow round hammer. This tool driven into the wood, removing a round plug about 12mm in dia. A fresh mush-room fruitbody was then squashed into the hole, and the bark allowed to settle back down over this inoculation to maintain moisture. Amazing that a technique developed a thousand years ago is still in use today. Similar to the old Chinese methods for cultivating wood decomposing mushrooms, some still practice natural log culture. This is an unreliable method that is unlikely to produce economically viable farms in most of the world. This is mostly due to the fact it is not possible to control the timing or the duration of the crop in natural log cultivation. The mushrooms fruit when the conditions are right, and the logs sit idle the rest of the year. This makes it impossible to market a year-round crop. While Asia developed methods to cultivate wood decomposing mushrooms, in Europe the secondary decomposers like Agaricus (button mushrooms) were more popularly gathered and eaten. The mushroom collectors knew that the mushrooms would reappear in the same fields year after year. Around the turn of the 19th century in France, it became popular to go to these fields, and take a shovel full of dirt where the mushrooms were fruiting. This was taken back to the farm, and applied to some substrate, usually in the barnyard. It didn’t take long to figure out that when this “soil spawn” was tossed into the corner of the barn onto the old, partially decomposed straw and hay from the horses, that was where the most mushrooms would grow. From this observation, creative farmers began to purposely compost their used horse bedding and using this material to create mushroom beds. These beds were inoculated with this wild soil spawn, and the button mushroom industry was born. LECTURE 04PLENARY18FUTURE OF MUSHROOM FARMING~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The harvesting of mushroom “soil spawn”, known then as Brick Spawn, became an industry, and by the turn of the 20th century this spawn was being exported from France to the rest of Europe and the United States, where a new and popular industry developed into large scale button mushroom farming. While this soil spawn represented a rather hit-or-miss approach to cultivation, it was the best available method at the time. As the understanding of microbiology advanced, and the microscope came into more widespread use, researchers saw that what the soil contained was the mycelium of the mushrooms. The mycelium from this soil was easy to isolate and grow on different media, which by 1930 lead to the development of Grain Spawn. This was one of the first real breakthrough in creating sustainable, reliable, and consistent mushroom cultivation. Sterile tissue culture came into widespread use by the end of World War II, and the grain spawn become more reliable and more widely available, with many spawn suppliers coming into the market and the number of mushrooms farms exploding. By 1980, there wereapproximately 1000 button mushroom farms in America.Through the late 1980’s and the 1990’s, large companies such as Campbells took over much of the mushroom industry. The smaller mom-and-pop farms grew less and less viable, since the larger farms could produce mushrooms so much more economically. By the early 2000’s, the American mushroom industry was down to just a few large companies. In my 45 years in the mushroom industry, I have seen farm after farm go under, due to competition from the large companies. Any button mushroom farm in America today that produces less than about 500,000 pounds per month has a hard time surviving in today’s competitive marketplace. The button mushroom industry by 2022 has become stagnant, with not many new Agaricus farms being built. So, what is the future of mushroom farming?The more recent methods of cultivation for these wood decomposing species is to sterilize small bags of sawdust with steam, and grow the mushrooms in controlled conditions to allow for a reliable year-round harvest. This is a well proven technology that works. But it is also a technology that is being replaced by a newer method. The reason it is being replaced by newer methods is the cost of equipment and energy continues to go up as time goes on. This makes mushroom cultivation less profitable, and the construction of new farms following this method more expensive every year, and therefore less attractive. To understand the future potential for mushroom farming with these newer methods, we need to unwrap the cultivation steps a bit. First and foremost, we need to understand that growing mush-rooms to the exclusion of all other organisms is a very unnatural process. Yet we need to grow our mushrooms in a true monoculture system to have reliable yields. LECTURE 04PLENARY19FUTURE OF MUSHROOM FARMING~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This requires two things: Sterilize the substrate to kill off all the potential competitors prior to inoculation with our target species and maintain sterility to the best of our ability throughout the entire process. Consider sterilization: To be “sterile” means all organisms are killed, nothing is left alive. In the early days of sterilization, circa 1850, it was found that exposing a material to steam under. But a lot of things have changed since 1850, including the development of many different processes to sterilize material. Like the use of microwaves, cold plasma, ionizing radiation, and chemical agents. In the end, sterile is sterile, and it makes no difference how this sterility is achieved. While any of these methods could work to sterilize substrate, most of them are even more expensive than using steam for sterilization. The exception is chemical sterilization. The use of chemicals to sterilize the substrate is termed the Cold System of Mushroom Cultivation. This is the primary type of mushroom farm being built today because the use of cheap and available chemicals is more profitable and less capital intensive then the use of steam. I predict the future of smaller scale, and probably larger scale mushroom farms, are purposedesigned, HEPA filtered sterile “Biological Manufacturing” facilities, using a cold sterilization process. We have come a long way from throwing a shovel full of dirt into the back corner of the barn. Todays and tomorrow’s farms are designed using an end-to-end systems approach of sterile manufacturing. For a mushroom farmer to try to compete in today’s market and with today’s ever-increasing costs of supplies and materials using yesterday’s technology doesn’t work very well. Tomorrow’s farms are more like a space station than a barn. Clean Rooms containing no wood, sterility as the number one operational feature, and a well-trained staff. This leads to a reliable and consistent yield, and an easy path to market and profitability.LECTURE 04PLENARY20MUSHROOMS AS SOURCE OF FUNCTIONAL FOOD INGREDIENTS~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Jovana Petrović~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Institute for Biological Research “Siniša Stanković”, University of Belgrade,National Institute of the Republic of Serbia, Bulevar despota Stefana 142, 11040 Belgrade~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Ever since the origin of the first man, food has been used not only to satisfy hunger, but to provide nutrition. However, it seems as thoughtful eating, with enjoying in food for nourishment is currently in the backgorund, and only hedonic experience of food in mind. Medicine and science revealed that the malnutrition on one side, and overeating on the other, are responsible for various diet relatedissues, such as: vitamins deficiency, scurvy, pellagra (in case poor diet) or coronary heart disease, high blood pressure, atherosclerosis, type 2 diabetes etc. (in case of excessive food intake). Nevertheless, the awareness on how eating habits influence health status has been set out as a trend only recently, opening up new avenues for the food industry development. This has been largely influenced by the increasing demand for healthy, nutritious and tasteful food of sustainable origin, which aside from fulfilling basic nutritional needs, also have health beneficial properties, thus fitting into the concept of functional food. Mushrooms emerged as rather versatile food which blend equally well with both sweet and savory ingredients, with nutritional richness, low caloric value, taste, and nutraceutical properties; their intake satisfies need for both nutrient and non-nutrient compounds that benefit human health in several aspects making them excellent candidates for functional food. Thus, regular consumption of mushrooms will provide recommended amount of fiber and other carbohydrates, proteins and contribute to daily intake of vitamins (B1, B2, B12, C, D, E), as well as polyunsatturated fatty acids. Along with this, it will supply with minerals (iron, copper, manganese, zinc) as well as terpenoids, phenolic and flavonoid compounds, which do not contribute to nutritive features of mushrooms, but play an important role in proper functioning of different metabolic pathways and/or have bioactive properties. Recent fundamental and clinical research highlighted potential of mushrooms as source of compounds with wide ranging bioactivities, including antimicrobial, antitumor, antiinflammatory, hepatoprotective, neuroprotective, cardioprotective, imunostimulatory etc. Palatability, versatility and bioactivity of wild growing and cultivated mushrooms with high demand of consumers for novel, tasteful and nutritive products, put them in the spotlight of the contemporary food industry, which led to the development of innovative designing mushroom-based food and beverages with functional food properties. Furthermore, extractability of bioactive/nutritive compounds from mushrooms enables their availability in form of pills/capsules which allows consumption in populations with limited accessibility to these highly valued ingredients. After all, since cultivation of mushrooms increases the amount of this underutilized food products and permits circular economy, it is expected that mushrooms and/or mushroom based food ingredients will soon be accessible to everyone.LECTURE 05PLENARY21ASPECTS IN MOREL MUSHROOMS QUALITIESAND HEALTH PROMOTING ACTIVITIES~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Segula Masaphy1,2~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~1 Applied Microbiology and Mycology Department, MIGAL, P.O. Box 831, 11016 Kiryat Shmona, Israel 2 Tel Hai College, 12210 Upper Galilee, Israel~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Morels (Morchella spp., Pezizales, Ascomycota) are edible mushrooms appreciated worldwide mainly for their taste and aroma. However, they are also rich in health promoting activities. These mushrooms have been in use in traditional medicine for centuries, especially in Southeast Asian counties. More recently, mushrooms of this genus became one of the most hunting targeted mushrooms by citizen in western countries as well. In nature, in most cases, they appear once at site, in unexpected manner. For many years, morels were difficult to cultivate due to lack of understanding of the conditions regulating their fruiting. In recent years, with the increasing of the knowledge on morels behavior in nature, there are increasing cases of successful cultivation practices of different species of morels, in outdoor or indoor cultivation systems. Yet, in most countries, cultivation of morels is still unstable, and the mushrooms are mostly harvested from the wild or their mycelium is grown in fermented culture, for consumption as a functional food and for food-flavoring. Morel species demonstrate high phenotypic as well as ecotyping diversities, resulting in different bioactivities. For example, the flavor’s molecules composition of morels was related to their ecotyping and phenotypic diversities. Similarly, the mushrooms in this genus have been found to contain high anti-oxidative activities, related to their phenotypic diversity. Moreover, the fruiting bodies were reported to harbor range of bacterial populations. This mushroom’s microbiome also affects the mushrooms metabolites production and bioactivities. Morels mushrooms demonstrate range of health promoting bioactivities as anti-inflammatory bioactivities, immunostimulatory and anti-tumor properties. These health benefits were attributed mainly to poly-saccharides as the active compounds, and to various phytochemicals, mainly phenolic compounds, tocopherols, ascorbic acid, and vitamin D. Morel’s nutritional composition was reported, including sugar, amino acid, fatty and organic acid, and mineral profile. The increasing in controlled cultivations systems might increase the control on the mushrooms qualities in all aspects, including their bioactive compounds production.LECTURE 06PLENARY22INTERACTIONS BETWEEN MEDICINAL MUSHROOM COMPONENTS AND CONVENTIONAL DRUGS – MOLECULAR BASIS AND PRACTICAL CONSEQUENCES~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Ulrike Lindequist~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~Institute of Pharmacy, Pharmaceutical Biology, University of Greifswald, D-17487 Greifswald, Germany; Email: lindequi@uni-greifswald.de~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Drug-drug interactions are a major issue in the application of drugs. Pharmacodynamic interactions take place when the pharmacological effect of one drug is altered by that of another drug or food component (Niu et al. 2019). Pharmacokinetic interactions occur when one drug interacts with another at the level of pharmacokinetics, i.e., absorption, metabolism, or excretion. Since the detection of the so-called grapefruit-effect and of interactions between St. John´s wort and conventional drugs some decades before we know that also interactions between herbal drugs and/or food and medicines can be relevant for the efficacy and/or safety of drugs, especially for those with a small therapeutic index.One underlying mechanism for pharmacokinetic interactions is the induction of cytochrome enzymes. Cytochrome P450 isoenzymes, especially CYP3A4, are responsible for the biotransformation of most xenobiotics including many drugs in clinical use. The other important mechanism is the influence on drug transporters. Uptake (members of SLC family) or efflux (members of ABC family, e.g., P-glycoprotein, ABCF1 = MDR1 and ABCC2 = MRP2) transporters facilitate the transport of drugs and influence drug exposure. In both mechanisms, the PXR receptor plays an important role. It is a member of the family of nuclear receptors and is involved in the regulation of metabolic processes in response to xenobiotics (Nicolussi 2019).There is some evidence of possible pharmacokinetic interactions caused by componentsof medicinal mushrooms. Examples include the influence of polysaccharides from Ganoderma lucidum on the activities of cytochrome P-450 in rat hepatic microsomes (Wang et al. 2007) and of Agaricus on this metabolizing enzyme also in vitro (Engdal and Nilson 2009). The relevance of these in vitro results for in vivo conditions remains unclear. A short time clinical study about the influence of Trametes versicolor on cytochrome P450 (Nicandro et al. 2007) did not reveal clinically relevant effects. The study of Toh et al. (2013) did show an effect of Lentinula edodes with a high content of ergothioneine on the renal clearance of gabapentine but did not estimate this to be clinically relevant.LECTURE 07PLENARY23INTERACTIONS BETWEEN MEDICINAL MUSHROOM COMPONENTS AND CONVENTIONAL DRUGS – MOLECULAR BASIS AND PRACTICAL CONSEQUENCES~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Pharmacodynamic interactions can be derived from the pharmacological profile of the respective partners. Possible examples are interactions between antidiabetics and hypoglycemic mushrooms or between anticoagulants and Auricularia auricula judae.The lecture explains the underlying mechanisms of interactions, gives an overview of the current state of knowledge on those originating from medicinal mushrooms, and provides information on possible consequences for their practical use. Literature: Engdal S, Nilsen OG (2009) Phytother Res 23(7): 906-912Nicandro JPA et al. (2007) J Herbal Pharmacother 7(1): 39-56Nicolussi S et al. (2019) Brit J Pharmacol 177: 1212-1226.Niu J et al. (2019) Clin Pharmacol Ther 105(6): 1395-1406.Toh DSL et al. 82013 Brit J Clin Pharmacol 78(1): 129-134Wang X et al. (2007) Biol Pharm Bull 30(9): 1702-1706Key words: medicinal mushroom, pharmacokinetic interactions, pharmacodynamic interactionsLECTURE 07PLENARY24ANTIFUNGAL AND ANTIOXIDANT POTENTIALOF SCHIZOPHYLLUM COMMUNE~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~~~~~Mikheil Asatiani*, Violeta Berikashvili, Tamar Khardziani, Aza Kobakhidze, Eka Metreveli, Eva Kachlishvili, Vladimir Elisashvili~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~The Institute of Microbial Biotechnology, Agricultural University of Georgia, Tbilisi, GeorgiaCorresponding author: m.asatiani@agruni.edu.ge~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Keywords: antifungal potential, higher basidiomycetes, plant pathogenic fungi, submerged fermentation, antioxidant activity.In the present research, twenty-three basidiomycetes belonging to different taxonomic groups were screened for their antifungal properties against plant pathogenic fungi, such as Aspergillus niger, Botrytis cinerea, Fusarium oxysporum, and Guignardia bidwellii. In the frame of the screening pro-gram several promising species with antifungal properties were detected. Among them Schizophyl-lum commune in submerged cultivation on glucose contained medium revealed the best antifungal potential (36%–66%). After optimization of the nutrient medium it was established that among differ-ent lignocellulosic substrates, the presence of 4% mandarin juice production waste (MJPW), caused the increase of antifungal activity (growth inhibition: A. niger – 65%, B. cinerea – 18%, F. oxysporum – 57%, G. bidwellii – 85%). Beside this, it was detected that 6% of peptone was a most appropriate nitrogen source to enhance antifungal properties of Sch. commune. Moreover, hot water, ethanol (80%), and ethyl acetate extracts obtained from submerged mycelium and culture liquid of Sch. commune contained considerable amount of bioactive substances such as, total phenolics, flavonoids and ascorbic acid. It is noteworthy that, ethyl acetate extracts obtained from the biomass and culture liquid of Sch. commune which was cultivated in presence of MJPW (6%) and peptone (6%) in nutrient medium, demonstrate highest content of total phenols, 120 GAE/g and 291 GAE/g, respectively. High content of ascorbic acid was observed in the same extracts(17 mg/g and 15 mg/g, respectively). In addition, the flavonoid content was considerable low to compare to the ethyl acetate extracts obtained from biomass and culture liquid after cultivation of Sch. commune on glucose conta"}}},{"rowIdx":83,"cells":{"Researcher Name":{"kind":"string","value":"Momoka Miyoshi"},"ORCID":{"kind":"string","value":"0009-0009-0454-1887"},"Researcher Query Keyword":{"kind":"string","value":"Self-Assembly of Titanate Nanosheets"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":84,"cells":{"Researcher Name":{"kind":"string","value":"Jixiu Niu"},"ORCID":{"kind":"string","value":"0009-0007-0041-9273"},"Researcher Query Keyword":{"kind":"string","value":"Allochroic Copper Iodide LEDs"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":85,"cells":{"Researcher Name":{"kind":"string","value":"David Grass"},"ORCID":{"kind":"string","value":"0000-0003-1781-0969"},"Researcher Query Keyword":{"kind":"string","value":"Pump-Probe Microscopy for Pigment Identification in Art"},"Research Full Paper":{"kind":"string","value":"{'Investigation of Artists Pigments with a Nonlinear Microscopy Technique': 'Title: Investigation of Artists Pigments with a Nonlinear Microscopy Technique\\\\nAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxii\\\\nCHAPTER\\\\n1. \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1\\\\n1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1\\\\n1.2 \\\\nIntroduction to NiTi shape memory alloy . . . . . . . . . . . . . . . 2\\\\n1.2.1 Applications of NiTi shape memory alloy . . . . . . . . . . 3\\\\n1.2.2 Martensitic phase transformation in NiTi . . . . . . . . . . 6\\\\n1.2.3 Functional and structural fatigue and fracture in NiTi . . . 8\\\\n1.2.4 Core: probing the role of phase transformation on SMA fa-\\\\ntigue and fracture with advanced techniques in experimental\\\\nmechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . 11\\\\n1.3 \\\\nIntroduction to digital image correlation . . . . . . . . . . . . . . . . 13\\\\n1.3.1 DIC fundamentals . . . . . . . . . . . . . . . . . . . . . . . 13\\\\n1.3.2 Types of DIC algorithms . . . . . . . . . . . . . . . . . . . 20\\\\n1.3.3 Speckle patterning . . . . . . . . . . . . . . . . . . . . . . . 21\\\\n1.3.4 Image capturing . . . . . . . . . . . . . . . . . . . . . . . . 23\\\\n1.3.5 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . 26\\\\n1.3.6 Validation and error evaluation . . . . . . . . . . . . . . . . 27\\\\n2. Cross polarization for improved optical digital image correlation . 29\\\\n2.1 \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29\\\\n2.2 Rigid body translation experiment . . . . . . . . . . . . . . . . . . . 32\\\\n2.2.1 Speckle patterning . . . . . . . . . . . . . . . . . . . . . . . 32\\\\nv\\\\n2.2.2 Imaging setup . . . . . . . . . . . . . . . . . . . . . . . . . 33\\\\n2.2.3 DIC calibration . . . . . . . . . . . . . . . . . . . . . . . . 35\\\\n2.2.4 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 35\\\\n2.2.5 DIC post-processing . . . . . . . . . . . . . . . . . . . . . . 36\\\\n2.2.6 \\\\nResults from rigid body translation experiment . . . . . . . 38\\\\n2.3 Uniaxial tension experiment . . . . . . . . . . . . . . . . . . . . . . . 44\\\\n2.3.1 Uniaxial specimen . . . . . . . . . . . . . . . . . . . . . . . 44\\\\n2.3.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . 45\\\\n2.3.3 DIC post-processing . . . . . . . . . . . . . . . . . . . . . . 46\\\\n2.3.4 Experimental \\\\nresults . . . . . . . . . . . . . . . . . . . . . 47\\\\n2.4 \\\\nDiscussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51\\\\n2.5 Summary and \\\\nConclusions . . . . . . . . . . . . . . . . . . . . . . . 53\\\\n3. Optimum paint sequence for speckle patterns in optical digital im-\\\\nage correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55\\\\n3.1 \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55\\\\n3.2 Observations and properties of paints . . . . . . . . . . . . . . . . . 56\\\\n3.3 DIC experiments and \\\\nresults . . . . . . . . . . . . . . . . . . . . . . 59\\\\n3.4 \\\\nConclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64\\\\n4. Residual stress and texture measurement of NiTi tube . . . . . . . . 65\\\\n4.1 \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65\\\\n4.1.1 Texture measurements . . . . . . . . . . . . . . . . . . . . 65\\\\n4.1.2 Notation conventions . . . . . . . . . . . . . . . . . . . . . 66\\\\n4.1.3 Tube textures . . . . . . . . . . . . . . . . . . . . . . . . . 66\\\\n4.1.4 Orientation distribution functions . . . . . . . . . . . . . . 66\\\\n4.1.5 Texture of NiTi tubes . . . . . . . . . . . . . . . . . . . . . 69\\\\n4.2 \\\\nMethods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70\\\\n4.2.1 Sample preparation . . . . . . . . . . . . . . . . . . . . . . 70\\\\n4.2.2 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . 73\\\\n4.3 \\\\nResults and \\\\ndiscussion . . . . . . . . . . . . . . . . . . . . . . . . . . 74\\\\n4.3.1 Pole figure measurements and orientation distribution func-\\\\ntion calculation . . . . . . . . . . . . . . . . . . . . . . . . 74\\\\n4.3.2 Residual stress analysis . . . . . . . . . . . . . . . . . . . . 81\\\\n4.3.3 X-ray penetration depth . . . . . . . . . . . . . . . . . . . 84\\\\n4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85\\\\n5. Grain size effects on NiTi shape memory alloy fatigue crack growth 86\\\\n5.1 \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86\\\\n5.1.1 Nanocrystalline NiTi shape memory alloys . . . . . . . . . 88\\\\n5.1.2 Fatigue crack characterization . . . . . . . . . . . . . . . . 89\\\\n5.2 \\\\nMethods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91\\\\nvi\\\\n5.2.1 Sample preparation . . . . . . . . . . . . . . . . . . . . . . 92\\\\n5.2.2 Macroscale fatigue measurements at high stress intensity . 94\\\\n5.2.3 SEM-DIC measurements . . . . . . . . . . . . . . . . . . . 96\\\\n5.2.4 Macroscale fatigue measurements at low stress intensity . . 98\\\\n5.3 \\\\nResults and \\\\ndiscussion . . . . . . . . . . . . . . . . . . . . . . . . . . 98\\\\n5.3.1 Microstructure . . . . . . . . . . . . . . . . . . . . . . . . . 99\\\\n5.3.2 Macroscale fatigue responses . . . . . . . . . . . . . . . . . 101\\\\n5.3.3 Microscale fatigue responses . . . . . . . . . . . . . . . . . 106\\\\n5.3.4 Fracture surfaces . . . . . . . . . . . . . . . . . . . . . . . 110\\\\n5.4 Summary and \\\\nconclusions . . . . . . . . . . . . . . . . . . . . . . . . 111\\\\n6. Texture dependence on functional and structural fatigue in NiTi\\\\nshape memory alloy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113\\\\n6.1 \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113\\\\n6.2 \\\\nIntroduction to texture dependence of NiTi in tension . . . . . . . . 113\\\\n6.3 \\\\nIntroduction to texture dependence of NiTi fatigue and fracture . . . 114\\\\n6.4 Materials and \\\\nmethods . . . . . . . . . . . . . . . . . . . . . . . . . 115\\\\n6.4.1 Superelastic NiTi sheet material . . . . . . . . . . . . . . . 115\\\\n6.4.2 Macroscopic tension experiments . . . . . . . . . . . . . . . 122\\\\n6.4.3 Macroscopic fatigue cracking experiments . . . . . . . . . . 124\\\\n6.4.4 Macroscopic digital image correlation . . . . . . . . . . . . 124\\\\n6.4.5 Microscopic SEM-DIC experiments . . . . . . . . . . . . . 125\\\\n6.5 \\\\nResults and \\\\ndiscussion . . . . . . . . . . . . . . . . . . . . . . . . . . 127\\\\n6.5.1 Cyclic uniaxial tension . . . . . . . . . . . . . . . . . . . . 127\\\\n6.5.2 Cyclic fatigue crack growth . . . . . . . . . . . . . . . . . . 131\\\\n6.5.3 Microscopic crack measurements . . . . . . . . . . . . . . . 134\\\\n6.6 Summary and \\\\nconclusions . . . . . . . . . . . . . . . . . . . . . . . . 139\\\\n7. The role of martensitic phase transformation on crack tip energy\\\\nin shape memory alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . 141\\\\n7.1 \\\\nIntroduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141\\\\n7.2 On the transformation toughening of NiTi during crack growth . . . 142\\\\n7.3 Background on fracture mechanics and crack opening displacement\\\\nmethods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144\\\\n7.4 Material and \\\\nmethods . . . . . . . . . . . . . . . . . . . . . . . . . . 146\\\\n7.4.1 Uniaxial tension . . . . . . . . . . . . . . . . . . . . . . . . 147\\\\n7.4.2 Crack tip opening displacements with constant force and\\\\nvaried temperature . . . . . . . . . . . . . . . . . . . . . . 148\\\\n7.4.3 Fatigue crack growth measurements . . . . . . . . . . . . . 149\\\\n7.5 \\\\nResults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150\\\\n7.6 \\\\nConclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156\\\\n8. \\\\nConclusions and future work . . . . . . . . . . . . . . . . . . . . . . . . 157\\\\nvii\\\\n8.1 \\\\nConclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157\\\\n8.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159\\\\n8.2.1 The role of inclusions on NiTi device life . . . . . . . . . . 161', 'Noninvasive identification of carbon-based black pigments with pump-probe microscopy': 'Title: Noninvasive identification of carbon-based black pigments with pump-probe microscopy\\\\nAbstract: Carbon-based black pigments, a widely used class of pigments, are difficult to differentiate with the noninvasive techniques currently used in cultural heritage science. We utilize pump-probe microscopy to distinguish four common carbon-based black pigments as pure pigments, as two-component black pigment mixtures, and as a mixture of a black and a colorful pigment. This work also demonstrates that even nominally “homogeneous” pigments present remarkable, and useful, heterogeneity in pump-probe microscopy. \\\\nIntroduction There is an unmet need in cultural heritage science for non-invasive identification of carbon-based black pigments, which are broadly used in paintings, drawings, and prints either by themselves or for shading another pigment (1). These pigments are easily produced through controlled burning of a material such as wood, bone, or oil (resulting in charcoal, bone black, and lamp black) or they occur naturally, such as graphite (2). They have been identified in some of the oldest pieces of art known to date, such as the cave paintings of Nawarla Gabarnmang in northern Australia (3). As their sourcing and cost are not prohibitive, carbon-based black pigments still represent one of the primary black pigment sources. There are two classification schemes for carbon-based black pigments (1, 2, 4-6). The first is by the level of order in the carbon network of the material, but this information is often inaccessible for black pigments incorporated into artwork. The second classification is based on the materials’ origin, but reliable information for pigments in historic works, particularly carbon-based based pigments where the naming conventions are tangled, is often missing. Material identification is essential for conservation of a work of art and provides insight into its historical context and provenance. In that respect, carbon-based blacks are problematic. The most specific method for identification of carbon-based black pigments, scanning electron microscopy with energy dispersive spectroscopy (SEM-EDX), has been used to distinguish carbon-based black pigments by morphology and, occasionally, elemental composition in pure reference samples (1, 2, 5, 7, 8). However, this requires physical removal of a cross-section from the work. Another approach is thermogravimetric analysis and differential scanning calorimetry. This study allowed for characterization of pure reference samples, but also requires invasive sampling (9). The go-to non-invasive \\\\nmethods in cultural heritage science are linear reflectance techniques, such as fiber-optic reflectance spectroscopy, hyperspectral imaging, multispectral imaging, and Raman spectroscopy due to their ease of use and portability (10-15). Unfortunately, linear reflectance measurements of carbon-based black pigments are featureless in the visible-NIR region (11, 12), and Raman spectroscopy is not well suited as the dominant spectral features of all carbon-based black pigments are the same. Presence of a carbon-based black pigment is confirmed by two characteristic peaks at approximately 1580 cm-1 and 1350 cm-1 (6, 16-19). The 1580 cm-1 peak (G Band) is the characteristic Raman peak for crystalline graphite (6, 16). The 1350 cm-1 peak (D or Disorder Band) is used as a measure of disorder in the carbonaceous material; it suggests the presence of heteroatoms in the graphitic structure, in-plane defects, or defects at the edge of the aromatic structure such as a tetrahedral carbon rather than the expected trigonal planar carbon (6, 16). Raman studies have attempted to delineate reference pigments by using the minute differences between these peaks, but so far, the \\\\nresults have not been encouraging (6, 17-19). Features in Fourier transform infrared (FTIR) spectra can be used to distinguish between reference carbon-based black pigments (5, 20). However, FTIR spectra from paintings have had mixed \\\\nresults. The spectra are either dominated by signals from the ground layer and the resin varnish; any features that would indicate a carbon-based pigment is overpowered by the other materials present (21) or they rely on other compounds present, like hydroxyapatite in ivory and bone black (22). X-ray fluorescence (XRF) is another noninvasive technique used in cultural heritage science; it cannot distinguish carbonaceous materials (23) but reveals secondary elements like Ca and P in compounds like hydroxyapatite to support an identification or rule them out (24-27). X-ray diffraction (XRD) can differentiate crystalline carbon-based black pigments (like graphite) from non-crystalline forms and can make further distinctions based on noncarbon components similar to XRF (1, 2, 5). However, it typically cannot differentiate non-crystalline, amorphous carbon-based black pigments (2, 4). Another study has shown good \\\\nresults in using powder XRD and a synchrotron beamline in identifying the type of carbon-based black pigment present in archeological samples, but required invasive sampling and powdering of the sample taken (28). Nonlinear optical microscopy techniques, such as two-photon fluorescence, second-harmonic generation, and coherent anti-Stokes Raman microscopy, have been shown to provide non-invasive, high-resolution imaging contrast in applications to biology (29) and, more recently, cultural heritage science (30-33). These contrasts are easily measured as they are emissive, generating light at wavelengths different from the excitation light. However, these conventional multiphoton techniques will not aid in distinguishing carbon-based black pigments; there is little to no fluorescence to analyze (12), and their Ramen spectra are not pigment specific (6, 16-19). We demonstrate here that another nonlinear optical technique, femtosecond pump-probe microscopy, can identify and distinguish common carbon-based black pigments non-invasively. Pump-probe microscopy takes advantage of the nonlinear interactions of two laser pulses with the sample to provide remarkable molecular specificity: in many cases there are multiple competing molecular mechanisms which provide significant contrast between nominally similar molecules. We focus here on transient absorption (TA), a subset of pump-probe, shown in figure 1, in which an excitation (pump) pulse affects the absorption of a time-delayed (probe) pulse. ‘Instantaneous’ mechanisms such as stimulated Raman scattering (SRS), two-photon absorption (SRS), sum-frequency generation (SFG), and cross-phase modulation (XPM) give signals only when pump and probe pulse overlap in time. Other molecular mechanisms result in delayed time signals. The pump laser pulse excites population into higher electronic states and simultaneously creates a population hole in the electronic ground state. Intermolecular vibrational redistribution rapidly rearranges the population of the electronically excited molecules, which can be transferred by the second pulse into a higher electronic state, namely excited state absorption (ESA) or to a vibrationally excited level of the ground state, namely stimulated emission (SE). The population hole in the ground state created by the pump pulse reduces the number of molecules available to be excited, reducing the absorption of the probe, a mechanism labeled ground state bleach (GSB). ESA and SE occur on roughly the same timescale, but the other effects have independent rates. Finally, pump absorption can cause localized heating, which in turn, can change the index of refraction. This affects, depending on the grain-size of the material, the Mie or Rayleigh scattering, effectively changing directional scattering, an effect called thermal scattering (TS). All of these non-emissive pump-probe interactions are separated from background signals using a modulation transfer scheme, explained in more detail in reference (34). Pump-probe microscopy has successfully been applied in a wide range of applications, including melanin characterization in biological tissue (35). This application provides a good example of the versatility: the melanin absorption spectrum is broad and featureless, but pump-probe images reveal significant heterogeneity from many competing molecular mechanisms shown in figure 1, and the contrast correlates with disease progression in melanoma. Previous cultural heritage applications include identification of iron oxides and red organic dyes, visualization of vermilion degradation, and as a tool to noninvasively obtain a virtual cross-section of historical works of art (36-41). Here, we use pump-probe microscopy to identify and distinguish four of the most used carbon-based black pigments, bone black, charcoal, graphite, and lamp black. We demonstrate that pump-probe microscopy reveals unique, nonlinear spectral features of these pigments that allow identification in two-compound black-black mixtures (which would be applicable in separating an underdrawing from black paint used in upper layers) and to identify black pigments in shading applications, i.e. a mixture of black with other colorful pigments, such as ultramarine blue. \\\\nResults Pump-probe spectroscopic features of black pigments: We acquired pump-probe (P-P) image stacks which are series of P-P images at 27 different time delays \\\\uf044t with a pump wavelength of lpump = 720 nm and a probe wavelength of lprobe = 817 nm. These stacks were taken from pure samples of bone black, charcoal, graphite, and lamp black. Each P-P stack consists of 128x128 pixels, and each pixel contains a P-P spectrum. These stacks of pure pigments were averaged across the spatial dimensions, that is over the imaged field of view, and normalized. The normalized spectra are shown in figure 2, while the non-normalized spectra can be found in the supplementary materials, figure S2. The P-P spectra of the four pigments exhibit distinctive qualitative differences. For graphite and lamp black, the duration of the temporal features is on the order of 100 fs, limited by the temporal resolution of our microscope. These ‘instantaneous’ signals suggest the involvement of virtual energy states in the nonlinear interaction, typical of processes like TPA or SRS. In our convention, transient loss processes such as TPA are depicted as positive while transient gain processes such as SRS, with the chosen pump-probe combination, are depicted as negative signals. This indicates TPA as the likely signal origin for graphite and lamp black. The spectra of bone black and charcoal are dominated by multiple excited state absorption processes, which are described by a superposition of multiple exponential decays. The obvious differences in the P-P spectra highlight the potential of P-P microscopy to noninvasively identify and distinguish these four carbon-based black pigments. Pure pigment heterogeneity: The pump-probe images uncover variations within pure pigments that are not apparent in their averaged spectra. For example, the averaged P-P spectrum of bone black is uniformly positive, peaking around a time delay of Dt = 0.5 ps. However, high-resolution pump-probe images, shown in figure S3 in the supplementary material, reveal interspersed regions of positive and negative signal. A convenient way of visualizing heterogeneity in pump-probe stacks is an adapted form of phasor analysis (42). In phasor analysis, single-frequency sine and cosine Fourier coefficients are calculated for each pixel and plotted as the 𝑥- and 𝑦- coordinates in a phasor histogram. For example, phasor coordinates of positive (negative) single-exponential decays would map onto a specific point Figure 1. Multiphoton nonlinear processes. Figure 2. Spatially averaged pump-probe spectra of bone black, charcoal, graphite, and lamp black. on the semi-circle in the first quadrant (third quadrant). Nearby points in a phasor diagram correspond to similar P-P signals. We show phasor histograms, computed with a frequency of f = 0.25 THz, of the pure black pigments in figure 3 (A, D, E, H). It is evident that the phasor histograms for bone black and charcoal fall into two distinct areas, aligning with positive and negative P-P spectra, respectively. Selecting the clusters indicated by red and yellow circles, we plot their corresponding P-P spectra in figure 3 (B, C, F, G) using the corresponding colors. For example, the phasor histogram of pure bone black is shown in figure 3A. We average all (pixel) spectra of the P-P image that fall into the colored circle in the phasor histogram and plot them in figure 3B in the corresponding color. The signals in both, bone black and charcoal, appear to be the same aside from a sign difference, respectively. This suggests TS as an underlying molecular mechanism. A pump-induced change in refractive index transiently changes the angular distribution of the backscattered light, that in combination with an aperture in the beam path causes a sign-change in the measured signal. An alternative interpretation would be the presence of two distinct chemical species in bone black and in charcoal. Investigation of the molecular origin of these signals goes beyond the scope of this manuscript as we found an effective approach to deal with this type of pigment heterogeneity. Conversely, graphite and lamp black appear homogeneous in their phasor histograms. Pigment assignment by unmixing: While the averaged spectra of the pigments shown in figure 2 are distinctive and distinguishable, these representations obscure the underlying heterogeneity within the pigments. For a typical cultural heritage application, a specific region or volume of interest would be imaged, and we would like to derive pigment identity in this specific region. This is a common problem in many applications such as hyperspectral imaging or Raman imaging, where the spectra of reference samples are well-known. The goal is to derive an abundance map from the hyperspectral or Raman image, determining the proportion of each reference spectra within every pixel. We apply the same principle to P-P images of black pigment mixtures. We use the averaged P-P curves shown in figure 3 as reference spectra. Due to the noted heterogeneity in bone black and charcoal, phasor analysis is utilized to derive two distinct reference curves for each of these pigments. Subsequently, in the unmixing process, probabilities for the two reference spectra associated with the same pigment are combined. We used a fully constrained least square algorithm based on reference (43) and implemented in pysptools 0.15.0 by Christian Therien. The model incorporates a non-negativity constraint, which permits only positive coefficients in constructing the measured signal as a weighted sum of reference spectra. Additionally, it enforces the constraint that the coefficients for a single pixel sum up to one. When presented with the black pigment mixtures, the unmixing approach correctly identified only 65% of the pixels in both the black-black mixtures and ultramarine blue-black mixtures. Given the low accuracy, we changed our approach to \\\\nmethods that are better suited to the complex and heterogeneous nature of the black pigment P-P data. Machine learning for pigment classification: Two popular \\\\nmethods have been used in the past to evaluate pump-probe spectra and assign them to molecular species: Principal component analysis and model fitting. Because several nonlinear optical interactions such as TPA, SRS, and ESA contribute to the measured pump-probe signals, the resulting P-P spectra are generally bipolar superpositions of multiple exponential decays and intrinsically non-orthogonal. As principal component analysis relies on orthogonal data structures, we consider it non-ideal for identification of carbon-based black pigments. Model fitting of P-P spectra with exponential basis functions is a powerful method and could allow for pigment identification based on specific lifetimes and amplitudes. However, there exists no method to unambiguously separate the superposition of multiple exponential decays into fundamental components. Also, there are limitations on how precise amplitudes and lifetimes from exponential decays can be extracted for a given signal-to-noise level (44, 45). Because the pump-probe signals from black pigments are generally weak, spatial resolution would need to be sacrificed by down-sampling to achieve an appropriate signal-to-noise level for fitting. For these reasons, and the intrinsic heterogeneity in the data, we decided to go a different route and train a support vector machine (SVM) for classification. An SVM is a supervised learning algorithm that, in its simplest form, classifies data into one of two classes. The algorithm takes n-dimensional input vectors (here pump-probe spectra consisting of 27 time delays) and separates them by a n-1 Figure 3. Phasor histograms and pump-probe signal components of the carbon-based black pigments. A, D, E, and H: The phasor coordinates of all signal-containing pixels in pure pigments (bone black, charcoal, graphite, and lamp black, respectively) as histograms. B, C, F, and G: Averaged P-P signals corresponding to the circled regions of the phasor histogram. dimensional hyperplane. The hyperplane, the plane that maximizes the margin between classes, is defined by the support vectors, the data points from each class that are nearest to the hyperplane and most influence its position. An SVM can be expanded to multiclass classification with a “one-versus-rest” strategy. It naturally lends itself to heterogeneous data and is well suited for high-dimensional data. We trained a SVM with P-P spectra from pure pigments and then used it to classify and identify pigments in two-component mixtures. Opposed to the unmixing approach, we train the SVM with around 6600 spectra of each of the five pigments (bone black, lamp black, charcoal, graphite, and ultramarine blue), thereby exposing the SVM to the full range of pigment heterogeneity. A description of the training, validation, and testing process can be found in the materials and \\\\nmethods section of this manuscript. In brief, the resulting SVM has an overall accuracy of 96% for pure pigments. This means that when the SVM is presented with a P-P spectrum from a single pixel of any of the five pigments (bone black, lamp black, charcoal, graphite, and ultramarine blue), it classifies it correctly 96 times out of 100. However, the performance of the SVM will drop for two-pigment mixtures to around 80%, as discussed in the next sections. Black-Black Mixtures: The accuracy reported above was obtained when identifying pure pigment samples. To investigate the SVM performance in a more relevant application, we tested it on six 50-50 (by weight) pigment mixtures of two different black pigments: bone black-charcoal, bone black-graphite, bone black-lamp black, charcoal-graphite, charcoal-lamp black, and graphite-lamp black. We took images of three different areas for each mixture and presented the P-P image stacks to the SVM classifier trained on pure pigments. We summarize the overall performance of the classifier with a bar chart in figure 4 one bar for each black-black mixture. It is crucial to note that we operate without a definitive \"ground truth\" in this context. While we know the two pigments comprising each mixture, their precise microscopic distribution remains unknown, and no alternative method exists for validating the P-P microscopy and classification \\\\nresults. To assess performance, we tally all classified pigments for a given mixture. We label pixels as correctly classified if they are part of the mixture and as misclassified if identified pigments are not part of the mixture. A complete table can be found in the supplementary information, table S1. A successful method to identify black pigments should have a high accuracy, meaning that the percentages of correctly identified pixels should be larger than any other misclassification. In our case the accuracy ranges between 62% (bone black-graphite) and 93% (bone black-lamp black). We will discuss the implications of these numbers in the \\\\ndiscussion section. The SVM returns a pigment classification for each pixel in the P-P stack, and we use this information to generate a false-color pigment map (or abundance map). Three representative examples are shown in figure 5. For the charcoal-graphite mixture in panel A, 71% of the signal containing pixels are identified as charcoal (39%) or graphite (32%). The remaining 29 % were misclassified, the worst offender being lamp black with 16 % of the pixels. The graphite-lamp black mixture, shown in figure 5B, performs well with 87% of the signal containing pixels being correctly identified (25% graphite, 62% lamp black). The largest misidentification is charcoal (11%). In the bone black-charcoal mixture, figure 5C, 87% of pixels are accurately identified and the largest offenders are graphite and lamp black, each with 7% of the pixels. While these samples were prepared as 50-50 mixtures by mass, each microscopic region in figure 5 deviates from that distribution. We do not expect to precisely measure the preparation ratio in a microscopic image of 36µm x 36µm. Figure 4. Summary of SVM performance on two-black pigment mixtures. The bar charts display the breakdown for classification of each black-black mixture. Each bar adds up to 100%. Green corresponds to correctly classified pixels. Red corresponds to misclassified pixels. Figure 5. Pigment map for three black-black mixtures. A: charcoal-graphite. B: graphite-lamp black. C: bone black-charcoal. Ultramarine Blue-Black Mixtures: In order to demonstrate P-P classification for shading applications, we imaged mixtures of each carbon-based black pigment with ultramarine blue (Na7Al6Si6O24S3), a modern equivalent to its natural form, lapis lazuli. P-P information for ultramarine blue is shown in the supplementary materials, figure S4. We captured images of twelve combinations, mixing each of the four black pigments with ultramarine blue in three mass ratios: 25:75, 50:50, and 75:25 percent. Brightfield images are in figure S5. For each of the twelve samples we imaged three different areas and computed the average classification accuracy. The effectiveness of the classifier across all the shading combinations is presented in figure 6, see table S2 with full details. Like in the black-black mixture case, there exists no ground-truth and we therefore apply the same metric to measure performance. Our methodology yields a robust qualitative classification, demonstrating a strong correlation between the detected amount of ultramarine blue and the actual physical mixing ratio of the sample. This correlation is consistent for most of the black pigments used. The classification accuracy ranges between 59% for the 75:25 lamp black-ultramarine blue mixture and 96% for the 50:50 bone black-ultramarine blue and 25:75 lamp black-ultramarine blue mixtures. Besides the two lowest performing instances (59% and 62%), the remaining 10 scenarios classify more than 70% correctly with a mean of 86%. The \\\\ndiscussion section will further explore the probable causes for this range in performance and suggest improvements. Analogous to the black-black mixtures, we can derive spatial maps of the black-blue mixtures. Three example maps of the mixtures, 75:25 ultramarine blue-charcoal, 75:25 ultramarine blue-graphite, and 50:50 ultramarine blue-bone black, are shown in figure 7. In all parts, the derived mixing ratios deviate from the real, macroscopic, mixing ratio. This highlights again the challenge in the absence of ground truth and the microscopic variation of mixing ratios. However, the images shown in figure 7, as well as the bar chart in figure 6, demonstrate good qualitative performance. \\\\nDiscussion Our findings highlight the potential of pump-probe microscopy for the noninvasive differentiation and mapping of black pigments. This technology presents a significant advancement in the noninvasive analysis of cultural heritage artifacts, where the precise identification of pigments can provide invaluable insights into the techniques and materials used by artists. However, there are cases of misclassification within the set of targets. Here we will discuss the performance and limitations of our current classification approach and then comment on the future application of pump-probe microscopy to more practical scenarios. Classification challenges and future strategies: Our current classification approach faces several challenges. The primary challenge we encounter with our current classification approach is the absence of a definitive ground truth for the images. The derived pigments maps, as shown in figure 5, are a result of P-P imaging and a SVM classifier. However, we know of no alternative \\\\nmethods to validate the correct mapping. The only ground truth we have is the mixing ratio used during sample preparation. The mixing ratio is a macroscopic quantity and the P-P images in this proof-of-principle study only sample three 36µm x 36µm areas. Within this small area, we expect variations in the pigment ratios that will vary from the macroscopic pigment distribution. For example, the measured pigment ratio in the 50:50 graphite-Figure 6. Summary of SVM performance on blue - black mixtures. A: bone black-ultramarine blue. B: charcoal-ultramarine blue. C: graphite-ultramarine blue. D: lamp black-ultramarine blue. Each bar adds up to 100%. Blue and green correspond to correctly classified pixels, blue for ultramarine blue and green for the black pigment. Red corresponds to the misclassified pixels. Figure 7. Pigment map for three ultramarine blue-black mixtures. A: 75:25 ultramarine blue-charcoal. B: 75:25 ultramarine blue-graphite. C: 50:50 ultramarine blue-bone black. lamp black mixture, in figure 5, is 25/62 \\\\uf0bb 0.4 which could be explained by a locally higher density of lamp black, rather than an inaccuracy in our method. A second challenge in pigment classification is the vast differences in signal-to-noise ratio (SNR) of different pigments. We show unnormalized curves of the four black pigments in the supplementary materials, figure S2. For example, typical bone black P-P signals are around 50 times weaker, and typical lamp black signals around 5 times weaker than signals of either charcoal or graphite. Signals of bone black and lamp black are close to the noise floor of our current microscope and are therefore more difficult to classify compared to the high SNR signals of charcoal and graphite. We believe that this is the main reason for the small bone black percentages identified in black-black mixtures, such as in figure 4. This also relates to the inherent tradeoff between spatial resolution and SNR: higher spatial resolution requires smaller sampling volumes and therefore leads to smaller SNR. Averaging neighboring pixels in an image increases SNR at the expense of spatial resolution. Spatial averaging, however, might mix spectra of adjacent pigments. This causes two additional challenges. In case of a large signal amplitude mismatch, the pigment with the larger signal will bury the spectral features of the weaker pigment signal thereby skewing classification towards pigments with larger signal. If spectra of comparable signal strengths are averaged, we create a de facto new spectrum that is unknown to the SVM classifier, which was only trained on pure spectra, leading to increased misclassification. Currently, we choose a compromise between spatial averaging to achieve sufficient SNR and maintaining enough resolution to resolve most individual pigment grains. In the future, we envision increasing SNR in two ways. First, by improving our detection capabilities, specifically by increasing collection efficiency by using higher numerical aperture objectives. Second, we intend to explore different pump and probe wavelengths that might offer a larger interaction cross-section for the currently weak signals of bone and lamp black. We also observe, at least heuristically, an increased misclassification rate at pigment grain boundaries. This can be seen in figure 7A and B, in which bone black appears around the edges of individual pigment grains. This can also be observed for other pigments such as lamp black. Whether this is caused by the grain boundary itself, by data thresholding, by the intrinsically weak SNR of bone and lamp black, or by the choice of classifier algorithm needs to be investigated in the future. An additional challenge is the intrinsic heterogeneity of pump-probe signals in black pigments. We hypothesize that two factors contribute to this heterogeneity: First, the inherent microscopic heterogeneity of charcoal and bone black. Of the four pigments, only graphite has an ordered molecular structure (sheets of sp2 hybridized carbon), consistent with the observed homogeneous signal in the phasor histogram in figure 3E. Lamp black, which undergoes a gas phase carbonization during production, is microscopically uniform (on the scale of the resolution of our microscope). Charcoal and bone black are derived from extremely heterogeneous base material, e.g. wood and bone, respectively, maintain a solid or liquid structure during carbonization, and therefore retain part of their initial structural complexity. The other factor that could contribute to the heterogeneity in pump-probe signal is the presence of heteroatoms or non-carbon constituents that commonly occur in carbon-based black pigments. Winter reports that incorporation of heteroatoms into the carbon matrix during the carbonization process is especially common for cokes and chars prepared at low temperatures (2). This may support the higher degree of heterogeneity of signals in charcoal and bone black. Heteroatoms can also be introduced via non-carbonaceous compounds that carry over from the source material. For example, bone black also contains the inorganic material hydroxyapatite, which would further explain why bone black displays more heterogeneity than charcoal. In our current classification scheme, we do not utilize local image information for classification. As of now, each pixel is classified independently from each other. However, given a specific pixel, there is a high probability that adjacent pixels are of the same pigments because pigments grains extend over multiple pixels. A logical next step is to explore classification with convolutional neural networks that are designed to respect local image information. Beyond proof-of-principle studies towards applications to works of art: This manuscript demonstrates the potential of P-P microscopy to noninvasively identify black pigments in mixtures. For this proof-of-principle demonstration we restricted ourselves to four black pigments and one colored pigment. Most works of art contain many more colors and, although we used the four most prevalent black pigments, there are other black pigments in use. Our group has analyzed a range of pigments, including red organic dyes, iron oxides, vermillion, and cadmium sulfide and we can add these pigments to our classification scheme (36-41). Furthermore, P-P microscopy offers two powerful and easily accessible degrees of freedom: the choice of pump and probe wavelength. Pump-probe signals reflect the population dynamics between molecular levels and are therefore strongly dependent on the pump and probe wavelengths. Pigments that present similar spectra at a particular wavelength combination may differ drastically at another (37). A convenient approach would be pigment exploration in a broadband pump-probe spectroscopy setup, where many wavelengths can be probed simultaneously. Then we would select a wavelength combination that offers a unique contrast for a specific pigment. Ultimately, multiple P-P images acquired with different wavelength combinations (hyperspectral pump-probe microscopy) will provide sufficient specificity to distinguish and identify many pigments. Extension of the SVM classifier to more pigments and to hyperspectral P-P images is conceptually straight forward and only requires the additional pure reference data in the training phase. In addition, polarization pump-probe microscopy, which offers improved chemically specific contrast based on the molecular anisotropy of pigments, can further improve pigment specificity (35, 46). The multiphoton nature of pump-probe microscopy enables high resolution in all three spatial dimensions, even beneath the surface of highly scattering materials (29). In previous experiments, we were able to image up to a depth of ≈ 90 𝜇𝑚 in paint layers to produce virtual cross sections (39), thus allowing cultural heritage scientists and conservators to better understand pieces of art without invasive sampling. However, achievable penetration depths depend on the absorption and scattering properties of the materials present at the surface and the subsequent layers. Carbon-based black pigments strongly absorb visible to near-infrared light and therefore reduce optical penetration depth. This will be most prominent in works of art with a thick or opaque layer of carbon-based black paint, for example in oil or tempera paintings. In a work with thinner or more transparent layers, such as watercolor paintings or drawings and prints, the absorption of the black pigments would not significantly reduce penetration depth. Our study has successfully shown that pump-probe microscopy is an effective noninvasive tool for differentiating black pigments in a variety of combinations, including mixtures with other blacks and with ultramarine blue. Across 18 different blends and shades, the technique accurately identifies 80% of the pixels in each image. This achievement highlights pump-probe microscopy\\\\'s capability to fill a significant void in the field of cultural heritage science, where, until now, no noninvasive method for identifying black pigments existed. We have outlined a clear strategy to further improve the performance and to increase the number of pigments in our approach and we envision applying this methodology to actual works of art. A particularly fascinating application would be Vermeer’s Girl with a Pearl Earring where bone black and charcoal are reported to exist together in an underlayer, currently only confirmed by analysis of a cross section (46). Pump-probe microscopy could be used to further validate these findings as well as to provide new information, i.e. a three-dimensional pigment map of both pigments across the painting. Materials and \\\\nMethods Pump-Probe Microscopy: A schematic of our pump-probe microscope is shown in figure S1. The output of a Ti:Sapphire laser (Coherent Chameleon II) with an 80 MHz repetition rate is split into two parts. One part serves as probe beam at a wavelength of \\\\uf06cprobe = 817 nm. The second part is frequency converted into the pump with a wavelength of 𝜆𝑝𝑢𝑚𝑝= 720 nm with an optical parametric oscillator (Coherent Mira-OPO). The pump pulse train is intensity-modulated by an acousto-optical modulator at a rate of 2 MHz. Both laser beams are spatially superimposed, sent into a laser scanning microscope, and focused onto the sample with a 20x 0.7 NA dry objective. The inter-pulse delay \\\\uf044t between pump and probe is controlled with a translation stage in the probe beam path. We utilize a modulation transfer scheme to detect the weak signals generated by the nonlinear interaction between pump, probe, and sample. As the nonlinear interaction transfers the pump modulation onto the probe pulse train, these changes in absorption in the probe pulse train are measured with a photodiode and a lock-in amplifier. For pigment imaging we use a pump and probe pulse intensity of I = 4.4x108 W/m2, (corresponding to 0.25 mW), and image an area of 36µm x 36µm for 27 time delays \\\\uf044t spanning -10 to 25 ps. Thus, the resulting data structure (image stack) is a 3-dimensional data cube with two spatial and one temporal dimension. Each pixel in the pump-probe stack represents a P-P spectrum, the change of absorption as a function inter-pulse delay \\\\uf044t. Preparation of Pigments: Pigments were thoroughly mixed with gum Arabic in a separate vessel to prepare a smooth, watercolor paint. This paint was applied to a commercially prepared canvas in two layers, allowing for drying in between. The pigments were commercially sourced from AGS Company (graphite), Coates Charcoal (charcoal), and Rublev Colours (bone black and lamp black). Data pre-processing of P-P image stacks: Raw P-P data is pre-processed before training and classification in the following steps: (1) Due to pump-leakage into the probe beam and potential long-lived (\\\\uf074 \\\\uf03e\\\\uf03e 12.5 ns, the time spacing between consecutive pulses) radiative states at the probe wavelength, we average three P-P images at negative time delays (\\\\uf044t = -10 ps, -5 ps, and -2.5ps) and subtract them from the entire P-P stack, thereby eliminating a constant offset in the data. (2) To increase SNR, we apply a spatial moving average filter of kernel size two. (3) A global intensity threshold is applied to all P-P stacks to discriminate noise from actual P-P signals. (4) We reduce the image size by down sampling of a factor two, consistent with the moving average filter. (5) For pure pigment images we apply their corresponding phasor masks, shown in Figure 2. For pigment mixtures, we apply a phasor mask that includes a combination of phasor masks from all individual pigments. Support Vector Machine (SVM): We use the scikit-learn 1.4 (47) and the imbalanced-learn 0.12.0 (48) python packages for training, validation and testing of the SVM. We randomly select around 6600 pure pigment, single pixel P-P spectra (imblearn RandomUnderSampler) for ultramarine blue, bone black, lamp black, charcoal, and graphite. We split them into training and testing set with a ratio of 3:1. We perform a hyperparameter optimization (scikit-learn GridSearchCV) of 𝐶 and 𝛾 for the SVM (scikit-learn SVC) with the training set, a stratified-5-fold strategy (scikit-learn StratifiedKFold), accuracy as scoring metric, and a standard scaler applied to all spectra (scikit-learn StandardScaler). The performance of the best classifier is inferred by measuring accuracy of the classifier applied on the test set. The entire procedure is repeated 5 times, and we compute the average and standard deviation over all five runs. The validation accuracy of accvalid = (96.32 \\\\uf0b1 0.08)% and the testing accuracy of acctest = (96.47 \\\\uf0b1 0.25)% are comparable and let us conclude that the classifier is well-trained. We then use the best hyperparameters and the entire data set from homogeneous pigment samples (5 x 6600 spectra) to train the final SVM classifier that is used to classify in two-pigment mixtures. Note that the final classifier is solely trained on homogeneous pigment data and has not been trained with any data from mixed samples.', 'Non-destructive three-dimensional imaging of artificially degraded CdS paints by pump-probe microscopy': \"Title: Non-destructive three-dimensional imaging of artificially degraded CdS paints by pump-probe microscopy\\\\nAbstract: Cadmium sulfide (CdS) pigments have degraded in several well-known paintings, but the mechanisms of degradation have yet to be fully understood. Traditional non-destructive analysis techniques primarily focus on macroscopic degradation, whereas microscopic information is typically obtained with invasive techniques that require sample removal. Here, we demonstrate the use of pump-probe microscopy to nondestructively visualize the three-dimensional structure and degradation progress of CdS pigments in oil paints. CdS pigments, reproduced following historical synthesis methods, were artificially aged by exposure to high relative humidity (RH) and ultraviolet (UV) light. Pump-probe microscopy was applied to track the degradation progress in single grains, and volumetric imaging revealed early CdS degradation of small particles and on the surface of large particles. This indicates that the particle dimension influences the extent and evolution of degradation of historical CdS. In addition, the pump-probe signal decrease in degraded CdS is observable before visible changes to the eye, demonstrating that pump-probe microscopy is a promising tool to detect early-stage degradation in artworks. The observed degradation by pump-probe microscopy occurred through the conversion from CdS into CdSO4·xH2O, verified by both FTIR (Fourier-transform infrared) and XPS (X-ray photoelectron spectroscopy) experimentIntroduction Cadmium sulfide (CdS)-based yellow pigments, also known as cadmium yellow, are a group of important inorganic pigments in art history.[1] The introduction of these pigments to artists’ palettes was followed by the improvement of industrial manufacturing in the 19th-20th century.[2] These pigments were favored by prominent artists, including Claude Monet,[3] Vincent van Gogh,[4] Edvard Munch,[5] Henri Matisse,[6] and Pablo Picasso,[7] due to their vivid colors and bright hues. However, many masterpieces by these artists have been found to suffer from CdS degradation such as fading, darkening, chalking, and flaking. The quality of historical CdS,[7] the presence of synthetic residues,[5] additives used by artists,[4] and environmental preservation conditions[8] can influence the degradation behaviors, complicating the conservation of artworks. Great efforts have been made by art conservators and researchers to understand the mechanisms underlying the deterioration. CdS is a semiconductor with a direct bandgap of 2.42 eV (512 nm) (energy diagram shown in Figure SI2).[9] Within the bandgap, deep trap states originate from CdS defects, such as surface sulfur and cadmium vacancies.[10] These trap states are suspected to drive CdS degradation, as reported in a recent study of Pablo Picasso’s Femme.[7] This effect was speculated to be more pronounced for small CdS pigments due to the increased surface-to-volume ratio.[7] In addition, photocatalytic activities of CdS nanoparticles with surrounding organic binders were suggested to be enhanced by an elevated number of defects, but further evidence is needed to fully understand the role of CdS particle size in the overall degradation process.[7] The synthesis method can affect the quality of Cd-based pigments and the stability of paints. Historical CdS pigments were mainly produced by either a dry or a wet method.[1] The dry method involved the calcination of metallic cadmium, cadmium oxide, or cadmium carbonate with a stoichiometric amount or an excess of sulfur at 300-500°C; the resulting CdS was washed and ground prior to applications in paintings. The wet process precipitated CdS from sulfide (hydrogen sulfide, sodium sulfide, or barium sulfide) and cadmium salt (cadmium chloride, cadmium nitrate, or cadmium sulfate) solutions; the precipitate was then washed and used without further thermal treatment. The pigments produced by wet methods are more prone to degradation due to their poor crystalline structures, smaller dimensions, and the presence of synthetic residues and/or byproducts from insufficient washing procedures.[11] For example, cadmium chloride (CdCl2) was detected in some extensively deteriorated sections in historical paintings.[6a-c, 12] However, the exact role of residues remains unclear, and additional studies are required to reveal their influences on CdS degradation. Environmental conditions strongly affect the degradation of CdS paints. In artificial aging experiments, humidity was identified to be the primary cause of CdS degradation, while light and elevated temperatures could exacerbate degradation,[8] consistent with the hypothesis that soluble impurities affect degradation mechanisms.[5] Traditional nondestructive techniques such as Raman, UV-Vis-NIR reflectance, X-ray fluorescence, and photoluminescence spectroscopy have been employed to study CdS degradation. However, due to their limitation in resolution, these techniques typically acquire only macroscale information. Synchrotron radiation-based X-ray spectro-microscopic methods[5] do provide high chemical and spatial resolution but are mostly restricted to analyzing cross-sectional samples. Here, we demonstrate the use of pump-probe microscopy, a nonlinear optical technique, to nondestructively generate three-dimensional, high-resolution maps of paint structures and to track the degradation process on a microscopic scale. Traditionally 2 used in biological imaging,[13] this technique has recently been applied in cultural heritage to create virtual cross-sections of paintings,[14] differentiate red organic dyes,[15] and identify the vermilion degradation product in a 14th-century painting.[16] The principle of pump-probe microscopy can be found in a previous review[17] and will only be briefly described here. The experimental setup is shown in Figure 1a. The pump-probe approach generates contrast by nonlinear optical interactions between the sample and two synchronized femtosecond pulse trains, named pump and probe. The pump pulse train is amplitude-modulated and superimposed with the probe pulse train before entering the laser scanning microscope. Pump and probe pulse trains interact nonlinearly with the sample in the focal region, thereby transferring some amplitude modulation from the pump to the probe pulse train (see Figure 1b). The backscattered probe light is collected while the pump light is removed with a filter, and the amplitude modulation of the probe pulse is measured with a photodiode and a lock-in amplifier. Changing the inter-pulse time delay (τ) between pump and probe results in characteristic transient absorption curves. In this study, we operate the microscope at a 720 nm pump and 817 nm probe wavelength combination with 0.4 mW of optical power in each pulse train. Many nonlinear interactions are accessible with pump-probe microscopy (see Figure SI3), but for our wavelength combination the predominant interaction between laser pulses and CdS is two-photon absorption (TPA, see Figure 1c). TPA involves a virtual energy state, and the sample absorbs one photon from each pulse only when pump and probe pulses arrive simultaneously at the sample (τ = 0 ps). The multiphoton nature of pump-probe microscopy allows for virtual sectioning and volumetric imaging in heterogeneous paint samples. By scanning the focus of the laser beams across the sample in both lateral directions and moving the sample in the axial direction, we can non-destructively map CdS grains within paints. This allows us to monitor CdS changes during artificial aging by recording the reduction of TPA signals. Figure 1. Pump-probe microscopy. a) Simplified schematic of the experimental setup. For details, see the Supporting Information. b) Modulation transfer scheme. c) Two-photon absorption (TPA). The dashed line indicates a virtual state. In this study, we present pump-probe imaging of artificially aged CdS paints which are prepared according to historical recipes. We demonstrate that pump-probe microscopy can noninvasively monitor the degradation inside single grains during artificial aging with micrometer-scale resolution. In addition, early-stage alterations on CdS mock-up paints can be recognized before visual changes are apparent, highlighting the potential of pump-probe microscopy to detect the degradation of seemingly well-preserved artworks. Results and Discussion Reproduction and Characterization of Historical CdS Pigments To simulate the degradation in historical paintings through artificial aging, CdS was reproduced following historical wet methods. CdS pigments synthesized from wet methods are more susceptible to degradation because of the poor crystalline structure and potential synthetic residues.[5, 11a] CdCl2 was chosen as a starting agent because chloride was found in degraded regions in multiple paintings,[5] making it one of the most plausible ingredient of the historical manufacturing process. Na2S was selected as sulfur source since previous research showed that CdS made from Na2S had similar morphology and photoluminescence properties as historical pigments.[18] This wet method reaction is described below. More details about the synthetic procedure can be found in the Supporting Information. 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶2 + 𝑁𝑁𝑁𝑁2𝑆𝑆 → 𝐶𝐶𝐶𝐶𝑆𝑆 + 2𝑁𝑁𝑁𝑁𝐶𝐶𝐶𝐶 CdS (Sigma Aldrich) was purchased as reference sample, and both synthesized and commercial CdS powders were first characterized and then prepared as oil paints to be artificially degraded. Figure 2 displays commercially available (top row) and synthesized (bottom row) CdS samples: powders (a,d), derived linseed oil paints (b,e), and microscopic images (c,f). The particle morphology and crystal forms were characterized by SEM-EDS (scanning electron microscopy) images (Figure 3) and X-ray diffraction (XRD) spectra (Figure 4). Figure 2. Commercial and synthesized CdS. Commercial CdS (Sigma Aldrich): a) powder, b) linseed oil paint, and c) microscopic image. Synthesized CdS: d) 3 powder, e) linseed oil paint, and f) microscopic image. Commercial CdS contains large, semitransparent crystals, while synthesized CdS displays smaller grains without clear crystalline structures. The SEM image of Figure 3a indicates that commercial CdS has a smooth surface with a well-defined crystalline structure in agreement with the microscopic photos. The EDS (energy disperse X-ray spectroscopy) spectrum of commercial CdS in Figure 3c shows a high purity of CdS with a minor amount of C (which may originate from sample processing). The synthesized CdS is composed of grains with a rough surface and nanometer-sized particles, as shown in Figure 3b. The corresponding EDS spectrum shows small contributions from C, O, Si (possibly comes from the imaging substrate), and negligible amount of Na (synthetic residue) in addition to the expected Cd and S. There was no detectable amount of Cl, which suggests an effective washing process in synthesis procedures. The magnified SEM image (shown in Figure SI3) reveals the existence of aggregated grains from nano-scale particles. The nano-dimensional and aggregated particles have a larger surface-to-volume ratio than commercial CdS grains, which could reduce the specular reflectance and lead to a darker color than commercial CdS. Figure 3. SEM-EDS analysis of commercial and synthesized CdS. SEM images of a) commercial and b) synthesized CdS. c) EDS spectra of commercial (blue) and synthesized (red) CdS. Both show the elemental components in the circled regions of their corresponding SEM images. Figure 4a shows XRD spectra of the two CdS powders with reference lines[19] for hexagonal and cubic CdS. The commercial CdS shows a clear hexagonal structure, while the synthesized CdS exhibits a more cubic and amorphous structure. The crystal structure difference can also contribute to the color difference of two CdS (where commercial CdS shows a lighter color).[20] In addition, the broad XRD peaks of synthesized CdS are a sign of imperfect crystal lattice and nanometric crystal size.[21] UV-Vis-NIR reflectance spectra shown in Figure SI5a also confirm the color difference between two CdS samples. The sigmoidal-shaped reflectance spectrum of synthesized CdS is another sign of poor crystal structure.[5, 22] The fluorescence spectra (shown in Figure SI5b) show that commercial CdS has a sharp band gap emission at 512 nm which is typical for bulk CdS. The synthesized CdS presents no band gap emission, similar to archived historical CdS pigments examined in a previous study.[2] Figure 4. XRD spectra of CdS and references (hexagonal and cubic) crystal structures. Commercial CdS shows strong hexagonal peaks, while synthesized CdS shows a cubic and amorphous structure. Artificial Aging of CdS Paints Artificial aging of both CdS paints was performed in a homemade aging chamber.[8] The CdS paints were applied on microscope slides, and aged by UV light (~400 nm) irradiation and exposure to high relative humidity (RH) levels (~75%) for four weeks. These samples were labeled as sampleUV+RH. Control experiments were also performed in which identical samples were exposed to only UV light (sampleUV) or only high RH levels (sampleRH) or left in a dark and dry environment (sampleN/A). Pump-probe microscopy was utilized to monitor the change in CdS paints during the four weeks of aging. The commercial CdS samples exposed to both UV light and high RH levels (samplecom) showed only negligible signs of degradation; therefore, we focus on the synthesized CdS sample for the following sections, unless otherwise noted. More details on sample preparation, aging chamber design, and aging process are documented in the Supporting Information. Table 1 presents photos of synthesized CdS paints before and after aging with light and high RH levels. The paints did not exhibit apparent signs of degradation during the first two weeks, but slight alterations, such as a global loss of surface oil gloss, became visible after four weeks. The samples exposed to only high RH levels retained their oil gloss, suggesting a lower degree of degradation in comparison to the UV-exposed sample. Table 1. Photographs of synthesized CdS oil paints before and after aging[a]. Aging conditions Unaged sample Aged for one week Aged for two weeks Aged for four weeks UV + high RH level High RH level 4 [a] SampleUV and sampleN/A are not shown here since they did not show visible signs of degradation over four weeks. This is consistent with pump-probe experiments discussed later. The UV-Vis-NIR reflectance spectra shown in Figure 5 agree with visual observations. The reflectance curve of sampleUV+RH exhibited a blue shift indicating a universal color change of the paint to lighter shades of yellow. Figure 5. UV-Vis-NIR reflectance spectra of sampleUV+RH before and after aging. Pump-probe Signatures of CdS Paints and Visualization of CdS Degradation To monitor the degradation progress on a microscopic scale, pump-probe microscopy was used to image CdS paints after various durations of artificial aging. Pump-probe images of the same region of interest (ROI) on sampleUV+RH before, after two weeks, and after four weeks of aging are shown in Figures 6a, b, and c. CdS grains are dominated by an instantaneous TPA signal, which can be seen in the transient absorption curves averaged over the rectangular areas plotted in Figure 6d. The signal decreases by 20%-30% during the first two weeks of aging and by more than 80% during weeks two to four. To map the change in pump-probe signals of CdS during aging, we co-registered pump-probe images of aged and unaged paints and computed the ratio of TPA signal between aged and unaged CdS. Figure 6e shows a pixel-by-pixel ratio image between two weeks of aging and no aging (Figures 6b divided by 6a). This ratio image presents the signal loss over two weeks of artificial aging. The false coloring is chosen such that unaltered regions are displayed in blue, and areas with drastically diminished signals are displayed in red. Green shows a moderate level of degradation. Similarly, Figure 6f shows the ratio image between four weeks of aging and two weeks of aging (Figure 6c divided by 6b). Beyond four weeks of aging, the pump-probe signal decreased to unobservable levels, limited by the signal-to-noise performance of our microscope. The pump-probe signatures of degradation are more severe in the second two weeks of aging (weeks two to four). Changes after two weeks were not yet observable by visual inspection or by bright field microscopy, but were obvious in the pump-probe data, emphasizing the ability of pump-probe to detect early-stage degradation. Figure 6. Pump-probe false-color images of sampleUV+RH a) before aging, b) aged for 2 weeks, and c) aged for 4 weeks d) Pump-probe transient absorption curves spatially averaged over the white rectangle in the pump-probe images. Pump-probe pixel-by-pixel ratio images of e) 2 weeks over 0 weeks of aging and f) 4 weeks over 2 weeks of aging. To visualize early-stage degradation in three dimensions on the surface and the interior of the paint, the sampleUV+RH was investigated after aging for one week (no degradation visible in color or gloss, neither by eye nor by bright field microscopy). Figure 7 shows a volume ratio image between one week of aging and no aging. A video of a full three-dimensional scan through this volume can be found in the Supporting Information. Small grains and the surface of large grains exhibited a moderate degree of CdS degradation shown in green. In addition, the inside of large grains stays unaffected, shown in blue. These observations indicate a particle size influence on the degradation of CdS: smaller particles experience more degradation in a shorter period of time. This result supports the previous hypothes"},"Similar Full Paper":{"kind":"string","value":"{'On-Chip Polarization Light Microscopy': 'Title: On-Chip Polarization Light Microscopy\\\\nABSTRACT: Optical metasurfaces revolutionized the approach to moulding the propagation of \\\\nlight by enabling simultaneous control of the light phase, momentum, amplitude and polarization. \\\\nThus, instantaneous spectropolarimetry became possible by conducting parallel intensity \\\\nmeasurements of differently diffracted optical beams. Various implementations of this very \\\\nimportant functionality have one feature in common - the determination of wavelength utilizes \\\\ndispersion of the diffraction angle, requiring tracking the diffracted beams in space. Realization of \\\\non-chip spectropolarimetry calls thereby for conceptually different approaches. In this work, we \\\\ndemonstrate that random nanoparticle arrays on metal surfaces, enabling strong multiple scattering \\\\nof surface plasmon polaritons (SPPs), produce upon illumination complicated SPP scattered \\\\npatterns, whose angular spectra are uniquely determined by the polarization and wavelength of \\\\nlight, representing thereby spectropolarimetric fingerprints. Using µm-sized circular arrays of \\\\nrandomly distributed nm-sized gold nanoparticles (density ~ 75 µm-2) fabricated on gold films, we \\\\nmeasure angular distributions of scattered SPP waves using the leakage radiation microscopy and \\\\n2 \\\\n \\\\nfind that the angular SPP spectra obtained for normally incident light beams different in wavelength \\\\nand/or polarization are distinctly different. Our approach allows one to realize on-chip \\\\nspectropolarimetry by fingerprinting using surface nanostructures fabricated with simple one-step \\\\nelectron-beam lithography. \\\\nKEYWORDS: Spectropolarimetry, SPP, plasmonic, leakage radiation microscopy \\\\n \\\\n3 \\\\n \\\\nLight-matter interactions using metal nanostructures have opened a way to the manipulation of \\\\nlight beyond the diffraction limit based on the propagation and localization of surface plasmon \\\\npolaritions (SPPs), hybrid waves involving electromagnetic fields in dielectrics and collective free-\\\\nelectron oscillations in metals, with the corresponding field termed plasmonics.1,2 With unique \\\\ncharacteristics featuring extremely broad bandwidth, strong field enhancements and subwavelength \\\\nconfinement of electromagnetic energy, plasmonics has not only given rise to various applications, \\\\nsuch as nonlinear signal conversion,3-5 light harvesting,6-8 lasering9-11 and sensing,12-15 but also \\\\noffered the opportunity to bridge the gap between photonics and electronics with miniature \\\\ndevices.16,17 Furthermore, specially designed metal surface nanostructures, termed plasmonic \\\\nmetasurfaces18, have also demonstrated their ability to function as ultrathin planar optical devices \\\\nand to control the amplitude, phase and polarization of light by arranging plasmonic resonant \\\\nantennas of specific size, shape and distribution.18-22 One of the most important and recent advances \\\\nin this direction concerns the development of spectropolarimeters,23-25 which enable simultaneous \\\\nmeasurements of the spectrum and state of polarization (SOP) of light and constitute a powerful \\\\nanalytic tool with capabilities far exceeding those of separate polarimeters and spectrometers. \\\\nMetasurfaces employed in these spectropolarimeters operate analogously to SOP-sensitive blazed \\\\ndiffraction gratings by directing different polarizations into spatially separated diffraction orders, \\\\nwhose diffraction angles are determined by the incident light wavelength. \\\\nWhile on-chip metasurface-only polarimeters allowing for the simultaneous determination of all \\\\nStokes parameters have also been demonstrated,26,27 on-chip spectropolarimetry with any of \\\\nconventional metasurfaces seems to represent mission impossible because of principal difficulties \\\\nin encoding two independent information (polarimetric and spectroscopic) channels into one-\\\\ndimensional parameter (polar angle of in-plane propagation) space. Considering this issue from the \\\\nviewpoint of surface wave excitation, the problem is that the conversion from incident propagating \\\\nwaves to surface modes requires phase matching along the surface wave propagation, which is \\\\ntypically achieved by using periodic structures, so that changing the wavelength influences only the \\\\n4 \\\\n \\\\ndiffraction efficiency and not the propagation direction. Realization of the on-chip \\\\nspectropolarimetry calls thereby for conceptually different (from gradient metasurfaces23-27) \\\\napproaches that would allow to realize the aforementioned encoding. \\\\nRandom (disordered) arrays of surface nanoparticles (NPs) represent a striking alternative to \\\\nperiodic arrays exploited in metasurfaces. It is worth mentioning that the term “plasmonic \\\\nmetasurfaces” was first introduced when considering SPP scattering phenomena by random and \\\\nperiodic nanostructures, because these could conveniently be described applying the concept of SPP \\\\neffective index inside nanostructured areas.28 Random metal nanostructures have been found to \\\\nexhibit versatile abilities of controlling light, such as subwavelength light localization29 and second-\\\\nharmonic enhancement through multiple SPP scattering,30 white light generation,31 SPP routing by \\\\nscattering-free channels,32,33 and multiple wavefront shaping.34 \\\\nUnderlying physical mechanisms involved in the aforementioned phenomena have often in \\\\ncommon strong multiple SPP scattering and localization initiated either by external illumination29,30 \\\\nor by launching an SPP beam towards a scattering NP array.32,33 In general, localization of light \\\\nhappens because of interference in recurrent multiple scattering in random media, resulting in \\\\ninhibition of light propagation when the scattering mean free path decreases below the light \\\\nwavelength.35 Since two-dimensional waves can be strongly localized with any degree of disorder \\\\n(in the absence of dissipation), the only condition for the strong SPP localization is that the SPP \\\\npropagation length should be much larger than the SPP localization length.29,32 In practice, this \\\\nimplies that, due to the exponential divergence of the localization length with the scattering mean-\\\\nfree path, the latter should be similar to or smaller than the SPP wavelength (l ≤ λSPP). In the \\\\nprevious experiments on SPP scattering by random gold NP arrays fabricated using electron-beam \\\\nlithography (EBL), typical values of the SPP scattering mean-free path at near-infrared wavelengths \\\\nwere l ~ 250 nm,33,36 indicating that the regime of strong multiple SPP scattering and localization \\\\nwas realized in these random nanostructures. \\\\n5 \\\\n \\\\nIn this work, we demonstrate that random nanoparticle arrays on metal surfaces, enabling strong \\\\nSPP multiple scattering, produce upon illumination complicated SPP scattered patterns, whose \\\\nangular spectra are uniquely determined by the polarization and wavelength of light, representing \\\\nthereby spectropolarimetric fingerprints. We fabricate dense µm-sized circular arrays of randomly \\\\ndistributed nm-sized gold NPs, which are known to ensure the regime of strong SPP multiple \\\\nscattering at near-infrared wavelengths,32, 33, 36 and observe SPP excitation and scattering produced \\\\nby these arrays upon illumination with normally incident light beams of different wavelengths \\\\nand/or SOPs. Using the leakage radiation microscopy (LRM) in the image and Fourier planes, 33 we \\\\nfind that the angular distributions of scattered SPP waves form very complicated irregular patterns \\\\nthat are distinctly different for different wavelengths and/or SOPs of the incident light. We relate \\\\nthis well-pronounced effect to far-field interference of SPP waves originating from uncorrelated \\\\n(and spatially localized) SPP sources generated upon illumination by strong SPP multiple \\\\nscattering.29 We further demonstrate that these visually different SPP scattering spectra can be \\\\nquantitatively discriminated using the correlation analysis. Numerical simulations of angular SPP \\\\nscattering spectra conducted using the previously developed self-consistent approach37,38 are found \\\\nin good qualitative agreement with experimental data, emphasizing the importance of realizing the \\\\nregime of strong multiple SPP scattering. The obtained results validate thereby the proposed \\\\nconcept of on-chip spectropolarimetry by fingerprinting based on SPP excitation and scattering by \\\\nrandom surface nanostructures. Although this approach requires calibration of fabricated \\\\nnanostructures (by measuring the angular spectra for various SOPs and wavelengths), i.e., \\\\nregistering fingerprints, the design and fabrication procedure is exceedingly simple, involving only \\\\ngeneration of NP random coordinates that are subsequently used in the straightforward one-step \\\\nEBL or nanoimprint lithography fabrication. Note that the comparison of an unknown SPP \\\\nscattering spectrum (i.e., obtained with an unknown SOP and wavelength) to the calibrated spectra \\\\nfrom the database (i.e., the fingerprint identification), can greatly be facilitated using the wavelet-\\\\nbased data processing.39 \\\\n6 \\\\n \\\\nRESULTS \\\\nLRM imaging of SPP scattering. The concept proposed for on-chip spectropolarimetry is based \\\\non the hypothesis that outgoing SPP waves, produced by strongly scattering random NP arrays upon \\\\nillumination, form very complicated irregular patterns (Figure 1a) that are distinctly different for \\\\nincident light beams different in wavelength and/or SOPs. In order to demonstrate the validity of \\\\nthis approach we fabricated dense (~ 75 µm-2) circularly shaped (with radii R from 1 to 5 µm) \\\\nrandom arrays of gold 60-nm-wide and 70-nm-high NPs atop a 70-nm-thick gold film supported by \\\\na 170-µm-thick silica glass substrate (Figure 1b). The fabrication procedure involved deposition of \\\\na gold film and one-step EBL fabrication of random NP arrays (see the Methods in Supporting \\\\nInformation). Although gold NPs were attempted to produce with diameters of 50 nm, the resulting \\\\nNPs were often slightly larger by up to 20 nm due to proximity effects causing aggregation as \\\\nobserved in the fabricated samples with scanning electron microscope (SEM) images (Figure 1c,d). \\\\nThe choice of NP size and density was motivated by the experimental results of our previous studies \\\\nof SPP scattering, localization and waveguiding in the EBL-fabricated random NP arrays.32,33,36 \\\\nThese parameters might not be optimum for the considered purpose but have already proven to be \\\\nsufficient for ensuring the regime of strong multiple SPP scattering and localization. \\\\nLRM imaging of the SPP excitation and scattering by the fabricated arrays was conducted using a \\\\ntunable Ti-Sapphire laser with the output range of 720–1000 nm, whose output beam was \\\\nhorizontally polarized while quarter-wave and half-wave plates were used along with two linear \\\\npolarizers to control the SOP of the beam incident on the random arrays (Figure S1). The direct and \\\\nFourier images of SPP scattering were captured by two infrared (IR) charge coupled devices \\\\n(CCDs) in both image (Figure 2a) and Fourier (Figure 2b) planes with the aperture stop introduced \\\\nin the Fourier plane center to remove the directly transmitted laser beam. We observed in the \\\\nFourier plane that the angular distributions of scattered SPP waves form very complicated irregular \\\\npatterns. Furthermore, as the random array diameter was increased, the angular SPP scattering \\\\n7 \\\\n \\\\npattern became progressively more complicated (Figure 3a-3h). We relate these features to the \\\\noccurrence (upon illumination of random NP arrays) of strong multiple SPP scattering and \\\\nlocalization resulting in the formation of uncorrelated and spatially localized (within the SPP \\\\nwavelength) SPP sources, whose strengths and phases are determined by the incident light SOP and \\\\nwavelength. The strongest influence on the SPP far-field scattering pattern and its fine structure is \\\\nexpected to originate from the SPP sources located along the outer perimeter of random NP arrays. \\\\nSince the smallest angular width of interference maxima in the far-field is determined by the SPP \\\\ndiffraction and estimated to be ~ λSPP/R, the total number of SPP rays (maxima in the angular SPP \\\\nscattering pattern) can be evaluated simply as N ~ 2πR/λSPP. It turned out that this formula describes \\\\nfairly well the experimental results obtained for random arrays with R = 1, 2, 3, and 5 µm, \\\\npredicting N ~ 8, 16, 24 and 39 with the numbers of observed (Figure 3i-3l) maxima being N = 9, \\\\n21, 25, and 32. In the remaining part of the paper, we mainly focus on the results obtained for the 3-\\\\nµm-radius random NP array that is already large enough to demonstrate excellent performance as \\\\nelaborated also when discussing the results obtained. \\\\nThe Fourier image (Figure 2b), representing a circle with the radius corresponding to the SPP \\\\nwave-vector magnitude, offers a convenient way for qualitatively characterizing the angular SPP \\\\nscattering spectra by averaging CCD pixel signals near the SPP circle along the same azimuthal \\\\nangle θ (Materials and methods). We observed that the angular SPP scattering spectra although \\\\nbeing very irregular were stable with respect to the beam size (once that was large enough) but \\\\nchanged drastically whenever the SOP (Figure 2c) and/or light wavelength was changed. These \\\\nunique SPP scattering spectra represent thereby spectropolarimetric fingerprints of the incident light \\\\nbeam and can conveniently be presented using multi-line coloured bar encoding (Figure 2d) that \\\\nfacilitates visual comparison of different SPP scattering spectra. \\\\nCorrelation of angular spectra for different SOPs and wavelengths. The SPP scattering \\\\nspectra obtained for different SOPs or light wavelengths were quantitatively compared by \\\\n8 \\\\n \\\\ncalculating the correlation coefficients Cij (Materials and methods). Considering four main SOPs, \\\\nwe found that the spectra corresponding to the orthogonal (linear or circular) polarizations are \\\\npractically uncorrelated: Cij < 0.2 (Figure 4). At the same time, the correlation coefficients between \\\\nthe spectra (obtained with differently sized random areas and at different wavelengths) \\\\ncorresponding to linear and circular polarizations were found to be dispersed around but close to 0.5 \\\\nwith the upper bound decreasing gradually for larger random areas (Table S1). This difference in \\\\nthe correlation of spectra corresponding to different SOPs can be related to the circumstance that \\\\nany circular polarization has one electrical field component identical to that of any linear \\\\npolarization (Supplementary Section 1). Taking into account that the (close to) identical angular \\\\nspectra should have the unitary (or very close to) correlation coefficient, we conclude that the \\\\npolarimetric fingerprints of the main SOPs are substantially different. This conclusion can also be \\\\nsupported by direct visual observations when comparing the corresponding rows in the angular SPP \\\\nscattering spectra (Figure 4). \\\\nAfter establishing the uniqueness of the angular spectra for substantially different SOPs, we \\\\ninvestigated the influence of the angle α between two linear polarizations by recording the angular \\\\nSPP scattering spectra with differently-sized random NP areas for different angles α with respect to \\\\nthe horizontal polarization. It is seen that angular dependences of the correlation coefficient become \\\\nclose to the analytical expression: C(α)=cos2α, once the random area size becomes sufficiently \\\\nlarge (Figure 5a). The correlation coefficient decreases thereby with increasing α in the same way \\\\nas the polarizer transmission when increasing the angle between the incident polarization and the \\\\npolarizer orientation. Considering highly irregular features of the angular SPP scattering spectra \\\\nused to calculate the correlation coefficient, this seems surprising but can be argued for in a \\\\nstraightforward manner as in the previous case of correlation between linear and circular \\\\npolarizations (Supplementary Section 1). Although direct comparison of two angular spectra \\\\nobtained with two different linear polarizations is not very sensitive to small angles between these \\\\n9 \\\\n \\\\npolarizations, one can definitely increase the accuracy of the polarization angle determination by \\\\ncomparing the angular spectrum of an unknown polarization with all angular spectra recorded for \\\\ndifferent polarization angles and fitting the corresponding correlation coefficients to cos2α \\\\ndependence. \\\\nWe further investigated correlation between angular SPP scattering spectra obtained for different \\\\nillumination wavelengths, observing a rapid decrease in the correlation coefficient for increasing \\\\ndifferences in wavelengths (Figure 5b). Following the same arguments regarding the occurrence of \\\\nuncorrelated and spatially localized SPP sources (producing angular scattering spectra), one can \\\\nreason that these sources are formed by multiple SPP scattering and interference within random \\\\ncavities, whose quality factors are of the order of LSPP/λ (neglecting the SPP dispersion) where LSPP \\\\nis the SPP propagation length. The wavelength selectivity can then be estimated as δλ~ λ2/LSPP, \\\\nwhich results in the wavelength sensitivity δλ ~ 27 nm, when using the following parameters:16 LSPP \\\\n~ 24 µm at λ ~ 800 nm. The sensitivity evaluated in this way is consistent with the experimental \\\\nresults (Figure 5b), representing essentially a ballpark estimation. It is clear that, also in this case, \\\\none can definitely increase the accuracy of the wavelength identification by comparing the angular \\\\nspectrum for an unknown wavelength with all angular spectra recorded for different wavelengths. \\\\nDISCUSSION \\\\nThe challenge of determining the wavelength and SOP of an optical beam with planar (on-chip) \\\\nconfigurations has been approached in our work by making use of strong multiple SPP scattering in \\\\nrandom NP arrays. In order to elucidate the importance of strong multiple SPP scattering, we \\\\nconducted numerical simulations using the previously developed self-consistent approach to \\\\ndescribe the multiple light scattering by NPs represented within the framework of point-dipole \\\\napproximation, with the Green dyadic being approximated by analytic expressions available for the \\\\nnear- and far-field regions.37,38 Although there are certain limitations to the accuracy of this \\\\napproach,40 the simulations were generally found to accurately reflect the main physical features in \\\\n10 \\\\n \\\\nvarious scattering phenomena.38 Likewise in this case, we have found in our simulations of SPP \\\\nscattering spectra features similar to the experimentally observed ones: influence of the scattering \\\\narea size, difference in the correlation coefficients when comparing linear and circularly polarized \\\\nincident beams, deterioration in the correlation for misaligned linear polarizations and different \\\\nwavelengths (Supplementary Section 2). Additionally, our simulations highlighted the importance \\\\nof realization of strong multiple SPP scattering for differentiating right and left circular \\\\npolarizations. It is clear that, when individual NPs are isotropic in the scattering plane, weak \\\\n(single) scattering cannot produce different angular SPP scattering spectra, whereas strong multiple \\\\nscattering couples many NPs that are sufficiently away from each other, introducing thereby \\\\nchirality in the scattering response (Figure S3 and Table S2). Overall, the obtained results validate \\\\nthe proposed concept of on-chip spectropolarimetry by fingerprinting based on SPP excitation and \\\\nscattering by random surface nanostructures. \\\\nThe operation principle introduced in our work requires careful calibration of a fabricated NP \\\\narray by measuring the angular SPP scattering spectra for various SOPs and wavelengths in a \\\\nprocedure analogous to the registration of fingerprints. The creation of a database of \\\\nspectropolarimetric fingerprints is a bit tedious but straightforward procedure, and the database \\\\nsearch for a match can be accomplished with the correlation analysis and facilitated using well-\\\\ndeveloped wavelet-based data processing techniques.39 On the positive side, the design and \\\\nfabrication procedure of a (strongly) scattering NP array is exceedingly simple, involving only \\\\ngeneration of random coordinates for (sufficiently large and densely packed32,33,36) NPs so that these \\\\ncoordinates can subsequently be used in the straightforward one-step EBL or nanoimprint \\\\nlithography fabrication. Note that measurements of the angular SPP scattering spectra to be used for \\\\nthe database do not impose stringent requirements on the incident laser beam size (it should be just \\\\nlarge enough) and angular alignment (normal incidence with one-degree accuracy is easy to realize \\\\nand sufficient for reproducible measurements). Moreover, gold nanostructures and films are known \\\\nto be stable in normal environment and suitable for long-term applications as noticed in the \\\\n11 \\\\n \\\\nexperiments using gold nanostructures and films for SPP localization29-33, gap-plasmon based \\\\ngradient metasurfaces20,25,28 and V-groove plasmonic waveguides11. Otherwise, it is also very \\\\nsimple to coat these structures with a thin layer of SiO2 for better protection. Furthermore, we \\\\nbelieve that this approach can also be applied to identification of optical fields with different orbital \\\\nangular momenta24 using the same principle, although in this case the alignment of the optical beam \\\\nto be examined becomes very important because of inhomogeneous spatial distributions of the \\\\nincident field phase and amplitudes. Taking into account explosively increasing interest in spin-\\\\norbit photonics,41 scattering of optical vortex beams by random nanostructures deserves special \\\\nstudies, promising new possibilities for spin-orbit optical interactions.42 Overall, the proposed \\\\nfingerprinting principle opens, in our opinion, a way to on-chip realization of general recognition of \\\\ncomplex optical field patterns with possible applications not only in the spectropolarimetry but also \\\\nwithin optical wavefront sensing, optical beam tracing and positioning (including beam \\\\nautofocusing) as well as environmental sensing, circumventing the usual requirements of elaborated \\\\ndesign and precise nanofabrication. \\\\nMethods \\\\nFabrication. Commercially available (18×18 mm2) 170-µm-thick silica glass substrates were \\\\ncleaned in an ultrasonic bath with distilled water, acetone and isopropyl alcohol, during 5 minutes \\\\nconsecutively, and covered with a 3-nm-thin layer of Ti by means of thermal evaporation in order \\\\nto improve the adhesion to gold. After depositing a 70-nm-thick gold film, a substrate was spin \\\\ncoated with poly-methyl-methacrylate (PMMA) (~ 350 nm thick) serving as a resist layer for the \\\\nEBL. Disordered patterns were designed using a simple MATLAB program based on “random” \\\\nfunction to generate random coordinates of points with the density of 75 µm-2 within a circle of a \\\\ngiven radius (ranging from 1 to 5 µm), discarding in the process of generation the points that were \\\\ntoo close to each other (i.e., requiring the minimum distance to be larger than 50 nm). Thus created \\\\n12 \\\\n \\\\nfile of random coordinates was used in the EBL system to conduct the PMMA film patterning. The \\\\nfinal fabrication step involved the deposition of a 70-nm-thick gold film and lift-off in acetone. \\\\nLRM imaging and data processing. Fourier images of SPP scattering (Figure 2b) were treated \\\\nusing a simple MATLAB program for computing the angular SPP intensity distributions. First, we \\\\nmanually found the origin (k = 0) and the radius (k = NSPP – the SPP effective index) corresponding \\\\nto a given Fourier image. The angular SPP scattering spectra were then computed with the step of \\\\none degree in the azimuthal angle θ (Figure 2b) by averaging the signals from three neighbour \\\\npixels along a given angle. Since, in the experiment, we only needed to rotate the wave plates and \\\\ntune the laser to change the SOP and wavelength of the incident light, the origin and radius in the \\\\nFourier images were typically the same for the same structure (the wavelength dispersion of SPP \\\\neffective index is negligibly small in the wavelength range used). This circumstance allowed us to \\\\nsimplify the image treatment procedure significantly. Once the SPP scattering spectra were \\\\ndetermined, resulting in ordered arrays of 360 values (different for different SOPs and wavelengths \\\\nof the incident light), the correlations between different spectra, Im(n)(θ), were calculated using \\\\nstandard formulae:43 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐(𝐼𝐼𝑚𝑚, 𝐼𝐼𝑛𝑛) = 𝑐𝑐𝑐𝑐𝑐𝑐(𝐼𝐼𝑚𝑚, 𝐼𝐼𝑛𝑛)/𝜎𝜎𝑚𝑚𝜎𝜎𝑛𝑛, with 𝑐𝑐𝑐𝑐𝑐𝑐(𝐼𝐼𝑚𝑚, 𝐼𝐼𝑛𝑛) = 〈[𝐼𝐼𝑚𝑚(𝜃𝜃)− 〈𝐼𝐼𝑚𝑚〉] ∙[𝐼𝐼𝑛𝑛(𝜃𝜃) − 〈𝐼𝐼𝑛𝑛〉]〉, 〈𝐼𝐼𝑚𝑚(𝑛𝑛)〉 = 1360∑ 𝐼𝐼𝑚𝑚(𝑛𝑛)(𝜃𝜃)360𝜃𝜃=1 , and 𝜎𝜎𝑚𝑚(𝑛𝑛) = �〈(𝐼𝐼𝑚𝑚(𝑛𝑛)(𝜃𝜃) − 〈𝐼𝐼𝑚𝑚(𝑛𝑛)〉)2〉. \\\\n \\\\nASSOCIATED CONTENT \\\\nSupporting Information \\\\nFigures S1 to S4, Table S1 to S3, calculations and simulations of angular SPP scattering spectra, \\\\nreference list (PDF). \\\\n \\\\nAUTHOR INFORMATION \\\\nCorresponding Authors *E-mail: yic@mci.sdu.dk; seib@mci.sdu.dk \\\\n13 \\\\n \\\\nORCID \\\\nYiting Chen: 0000-0002-4003-4796 \\\\nSergey I. Bozhevolnyi: 0000-0002-0393-4859 \\\\nNotes The authors declare no competing financial interest. \\\\nACKNOWLEDGEMENT \\\\nThe authors gratefully acknowledge financial support from the European Research council, Grant \\\\n341054 (PLAQNAP) and the University of Southern Denmark (SDU 2020). \\\\nREFERENCES \\\\n1. Schuller, J. A.; Barnard, E. S.; Cai, W.; Jun, Y. C.; White, J. S.; Brongersma, M. L. \\\\nBrongersma, Plasmonics for extreme light concentration and manipulation. Nature Mater. \\\\n2010, 9, 193–204. \\\\n2. Gramotnev, D. K.; Bozhevolnyi, S. I. Plasmonics beyond the diffraction limit. Nat. Photon. \\\\n2010, 4, 83–91. \\\\n3. Mühlschlegel, P.; Eisler, H.-J.; Martin, O. J. F.; Hecht, B.; Pohl, D. W. Resonant optical \\\\nantennas. Science 2005, 308, 1607–1609. \\\\n4. Danckwerts, M.; Novotny, L. Optical frequency mixing at coupled gold nanoparticles. Phys. \\\\nRev. Lett. 2007, 98, 026104. \\\\n5. Kim, S.; Jin, J.; Kim, Y. J.; Park, I. Y.; Y. Kim, Kim, S. W. High-harmonic generation by \\\\nresonant plasmon field enhancement. Nature 453, 757–760 (2008). \\\\n6. Atwater, H. A.; Polman, A. Plasmonics for improved photovoltaic devices. Nat. Mater. \\\\n2010, 9, 205–213. \\\\n7. Landy, N. I.; Sajuyigbe, S.; Mock, J. J.; Smith, D. R.; Padilla, W. J. Perfect metamaterial \\\\nabsorber. Phys. Rev. Lett. 2008, 100, 207402. \\\\n14 \\\\n \\\\n8. Hao, J.; Wang, J.; Liu, X.; Padilla, W. J.; Zhou, L.; Min, Q. High performance optical \\\\nabsorber based on a plasmonic metamaterial. Appl. Phys. Lett. 2010, 96, 251104. \\\\n9. Oulton, R. F.; Sorger, V. J.; Zentgraf, T.; Ma, R. M.; Gladden, C.; Dai, L.; Bartal, G.; \\\\nZhang, X. Plasmon lasers at deep subwavelength scale. Nature 2009, 461, 629. \\\\n10. Shi, C.; Soltani, S.; Armani, A. M. Gold nanorod plasmonic upconversion microlaser. Nano \\\\nLett. 2013, 13, 5827–5831. \\\\n11. Bermúdez-Ureña, E.; Tutuncuoglu, G.; Cuerda, J.; Smith, C. L. C.; Bravo-Abad, J.; \\\\nBozhevolnyi, S. I.; Fontcuberta i Morral, A.; García-Vidal, F. J.; Quidant, R. Plasmonic \\\\nwaveguide-integrated nanowire laser. Nano Letters 2017, 17, 747–754. \\\\n12. Anker, J. N.; Hall, W. P.; Lyandres, O.; Shah, N. C.; Zhao, J.; Duyne, R. P. V. Biosensing \\\\nwith plasmonic nanosensors. Nat. Mater. 2008, 7, 442–453. \\\\n13. Kubo, W.; Fujikawa, S. Au double nanopillars with nanogap for plasmonic sensor. Nano \\\\nLett. 2011, 11, 8–15. \\\\n14. Mesch, M.; Metzger, B.; Hentschel, M.; Giessen H. Nonlinear plasmonic sensing. Nano \\\\nLett. 2016, 16, 3155–3159. \\\\n15. Kinkhabwala, A.; Yu, Z.; Fan, S.; Avlasevich, Y.; Müllen, K.; Moerner, W. E. Large single-\\\\nmolecule fluorescence enhancements produced by a bowtie nanoantenna. Nat. Photon. 2009, \\\\n3, 654–657. \\\\n16. Papaioannou, S.; Kalavrouziotis, D.; Vyrsokinos, K.; Weeber, J.-C.; Hassan, K.; Markey, L.; \\\\nDereux, A.; Kumar, A.; Bozhevolnyi, S. I.; Baus, M.; Tekin, T.; Apostolopoulos, D.; \\\\nAvramopoulos, H.; Pleros, N. Active plasmonics in wdm traffic switching applications. Sci. \\\\nRep. 2012, 2, 652. \\\\n17. Haffner, C.; Heni, W.; Fedoryshyn, Y.; Niegemann, J.; Melikyan, A.; Elder, D. L.; Baeuerle, \\\\nB.; Salamin, Y.; Josten, A.; Koch, U.; Hoessbacher, C.; Ducry, F.; Juchli, L.; Emboras, \\\\nA.; Hillerkuss, D.; Kohl, M.; Dalton, L. R.; Hafner, C.; Leuthold, J. All-plasmonic Mach-\\\\nZehnder modulator enabling optical high-speed communication at the microscale. Nat. \\\\n15 \\\\n \\\\nPhoton. 2015, 9, 525–528. \\\\n18. Zhao, Y.; Alù, A. Manipulating light polarization with ultrathin plasmonic metasurfaces. \\\\nPhys. Rev. B 2011, 84, 205428. \\\\n19. Yu, N.; Genevet, P.; Kats, M. A.; Aieta, F.; Tetienne, J. P.; Capasso, F.; Gaburro, Z. Light \\\\npropagation with phase discontinuities: generalized laws of reflection and refraction. \\\\nScience 2011, 334, 333–337. \\\\n20. Pors, A.; Albrektsen, O.; Radko, I. P.; Bozhevolnyi, S. I. Gap plasmon-based metasurfaces \\\\nfor total control of reflected light. Sci. Rep. 2013, 3, 2155. \\\\n21. Minovich, A. E.; Peter, M.; Bleckmann, F.; Becker, M.; Linden, S.; Zayats, A. V. Reflective \\\\nmetasurfaces for incoherent light to bring computer graphics tricks to optical systems. Nano \\\\nLett. 2017, 17, 4189–4193. \\\\n22. Kildishev, A. V.; Boltasseva, A.; Shalaev, V. M. Planar photonics with metasurfaces. \\\\nScience 2013, 339, 1232009. \\\\n23. Chen, W. T.; Török, P.; Foreman, M. R.; Liao, C.; Tsai, W.; Wu, P. R.; Tsai, D. P. \\\\nIntegrated plasmonic metasurfaces for spectropolarimetry. Nanotechnology 2016, 27, \\\\n224002. \\\\n24. Maguid, E.; Yulevich, I.; Veksler, D.; Kleiner, V.; Brongersma, M. L.; Hasman, E. Photonic \\\\nspin-controlled multifunctional shared-aperture antenna array. Science 2016, 352, 1202–\\\\n1206. \\\\n25. Ding, F.; Pors, A.; Chen, Y.; Zenin, V. A.; Bozhevolnyi, S. I. Beam-size-invariant \\\\nspectropolarimeters using gap-plasmon metasurfaces. ACS Photonics 2017, 4, 943−949. \\\\n26. Mueller, J. P. B.; Leosson, K.; Capasso, F. Ultracompact metasurface in-line polarimeter. \\\\nOptica 2016, 3, 42−47. \\\\n27. Pors, A.; Bozhevolnyi, S. I. Waveguide metacouplers for in-plane polarimetry. Phys. Rev. \\\\nAppl. 2016, 5, 064015. \\\\n28. Radko, I. P.; Volkov, V. S.; Beermann, J.; Evlyukhin, A. B.; Søndergaard, T.; Boltasseva, \\\\n16 \\\\n \\\\nA.; Bozhevolnyi, S. I. Plasmonic metasurfaces for waveguiding and field enhancement. \\\\nLaser & Photon. Rev. 2009, 3, 575–590. \\\\n29. Bozhevolnyi, S. I. Localization phenomena in elastic surface-polariton scattering caused by \\\\nsurface roughness. Phys. Rev. B 1996, 54, 8177–8185. \\\\n30. Bozhevolnyi, S. I.; Beermann, J.; Coello, V. Direct observation of localized second-\\\\nharmonic enhancement in random metal nanostructures. Phys. Rev. Lett. 2003, 90, 197403. \\\\n31. Novikov, S. M.; Frydendahl, C.; Beermann, J.; Zenin, V. A.; Stenger, N.; Coello, V.; \\\\nMortensen, N. A.; Bozhevolnyi, S. I. White light generation and anisotropic damage in gold \\\\nfilms near percolation threshold. ACS Photonics 2017, 4, 1207–1215. \\\\n32. Bozhevolnyi, S. I.; Volkov, V. S.; Leosson, K. Localization and waveguiding of surface \\\\nplasmon polaritons in random nanostructures. Phys. Rev. Lett. 2002, 89, 186801. \\\\n33. Pisano, E.; Coello, V.; Garcia-Ortiz, C. E; Chen, Y.; Beermann, J.; Bozhevolnyi, S. I. \\\\nPlasmonic channel waveguides in random arrays of metallic nanoparticles. Opt. Express \\\\n2016, 24, 17080–17089. \\\\n34. Veksler, D.; Maguid, E.; Shitrit, N.; Ozeri, D.; Kleiner, V.; Hasman, E. Multiple wavefront \\\\nshaping by metasurface based on mixed random antenna groups. ACS Photonics 2015, 2, \\\\n661–667. \\\\n35. Lagendijk, A.; Tiggelen, B. van; Wiersma, D. S. Fifty years of Anderson localization. Phys. \\\\nToday 2009, 62, 24-29. \\\\n36. Bozhevolnyi, S. I.; Volkov, V. S.; Leosson, K.; Boltasseva, A. Surface plasmon polariton \\\\nwaveguiding in random surface nanostructures. J. Microscopy 2003, 209, 209–213. \\\\n37. Søndergaard, T.; Bozhevolnyi, S. I. Vectorial model for multiple scattering by surface \\\\nnanoparticles via surface polariton-to-polariton interactions. Phys. Rev. B 2003, 67, 165405. \\\\n38. Beermann, J.; Bozhevolnyi, S. I.; Coello, V. Modelling of nonlinear microscopy of localized \\\\nfield enhancements in random metal nanostructures. Phys. Rev. B 2006, 73, 115408. \\\\n39. Mallat, S. A wavelet tour of signal processing; Academic Press, 1999. \\\\n17 \\\\n \\\\n40. Evlyukhin, A. B.; Bozhevolnyi, S. I. Point-dipole approximation for surface plasmon \\\\npolariton scattering: implications and limitations. Phys. Rev. B 2005, 71, 134304. \\\\n41. Cardano, F.; Marrucci, L. Marrucci, Spin-orbit photonics. Nat. Photon. 2015, 9, 776–778. \\\\n42. Bliokh, K. Y.; Rodríguez-Fortuño, F. J.; Zayats, A. V. Spin-orbit interactions of light. Nat. \\\\nPhoton. 2015, 9, 796–808. \\\\n43. Mc Quarrie, D. A. Mathemathical Methods for Scientists and Engineers, University Science \\\\nBooks, 2003. \\\\n \\\\n18 \\\\n \\\\nFigures \\\\n \\\\n \\\\nFigure 1. (a) Schematic of SPP excitation and scattering occurring upon illumination of random NP \\\\narrays at normal incidence. (b) Sketch of an individual gold NP atop a 70-µm-thick gold film \\\\ndeposited on a silica wafer. (c) Scanning electron microscope (SEM) image of a fabricated random \\\\nNP array with the radius of 2 µm. Scale bar, 0.5 µm. (d) Enlarged part of the SEM image shown in \\\\nc. Scale bar, 200 nm. \\\\n \\\\n19 \\\\n \\\\n \\\\nFigure 2. (a) LRM image of the SPP excitation and scattering in the direct (sample surface) plane \\\\nobtained with a random 5-µm-radius array being illuminated by a linearly polarized (in the \\\\nhorizontal direction) laser beam at the wavelength of 800 nm. (b) The corresponding LRM image in \\\\nthe Fourier plane introducing the azimuthal angle θ that denotes the SPP scattering direction. The \\\\nbright circle is formed by SPP waves scattered in different directions with its radius being related to \\\\nthe SPP effective index. (c) Angular SPP scattering spectra recorded with four different SOPs of \\\\nincident light at 800 nm, which are, from bottom to top, vertical polarization (black), horizontal \\\\npolarization (red), left-circular polarization (blue) and right-circular polarization (magenta). (d) The \\\\nSPP scattering spectrum obtained with the vertical polarization (shown also in c) and represented, \\\\nfor convenience in visual comparison of SPP spectra, using 8 coloured stripes (barcodes) displaying \\\\nsequentially, from left to right and from top to bottom, 45°-wide angular distributions. \\\\n \\\\n20 \\\\n \\\\n \\\\nFigure 3. LRM characterization of differently-sized random areas. a – d, LRM images of scattered \\\\nSPP waves obtained in the (direct) image plane with linearly polarized (in the horizontal direction) \\\\nlight at the illumination wavelength of 800 nm that is normally incident on differently-sized random \\\\nNP arrays with radii R = a, 1, b, 2, c, 3 and d, 5 µm, respectively. e – h, The corresponding images \\\\nobtained in the Fourier plane with the respective random areas and with the transmitted beam spot \\\\nbeing blocked. i – l, The corresponding angular SPP intensity distributions retrieved from the \\\\nrespective Fourier images. The numbers of distinct maxima in these distributions are identified as 9, \\\\n21, 25, and 32, respectively, by using a criterion that a local maximum should be situated between \\\\ntwo local minima with signal drops of more than 50% with respect to the maximum value. \\\\n \\\\n \\\\n21 \\\\n \\\\n \\\\nFigure 4. Multi-line barcodes representing angular SPP scattering spectra obtained with a random \\\\n5-µm-radius array being illuminated at the wavelength of 800 nm with (a) horizontal linear, (b) \\\\nvertical linear, (c) left-circular and (d) right-circular polarizations. Cmn indicate the corresponding \\\\ncorrelation coefficients between the spectra displayed. \\\\n \\\\n \\\\n \\\\n22 \\\\n \\\\n \\\\nFigure 5. (a) Correlation between SPP scattering spectra obtained for different linear polarizations \\\\n(at 800 nm) with differently sized random NP arrays as a function of the angle between the \\\\npolarizations. The qualitatively expected dependence, cos2α, is also indicated for comparison. (b) \\\\nCorrelation between SPP scattering spectra obtained for different incident light wavelengths with \\\\nlinear horizontal polarization as a function of the wavelength, when using different reference \\\\nwavelengths, including 750, 800, 850 and 900 nm. \\\\n', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mst"}}},{"rowIdx":86,"cells":{"Researcher Name":{"kind":"string","value":"Liting Tao"},"ORCID":{"kind":"string","value":"0000-0001-7278-1115"},"Researcher Query Keyword":{"kind":"string","value":"Vapor Deposition of Perovskite Heterojunctions for X-ray Detection"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Phase Modulation on Low‐Dimensional Perovskite for High Performance Blue Light Emitting Diodes': \"Title: Phase Modulation on Low‐Dimensional Perovskite for High Performance Blue Light Emitting Diodes\\\\n 저 시-비 리- 경 지 2.0 한민 는 아래 조건 르는 경 에 한하여 게 l 저 물 복제, 포, 전송, 전시, 공연 송할 수 습니다. 다 과 같 조건 라야 합니다: l 하는, 저 물 나 포 경 , 저 물에 적 된 허락조건 명확하게 나타내어야 합니다. l 저 터 허가를 면 러한 조건들 적 되지 않습니다. 저 에 른 리는 내 에 하여 향 지 않습니다. 것 허락규약(Legal Code) 해하 쉽게 약한 것 니다. Disclaimer 저 시. 하는 원저 를 시하여야 합니다. 비 리. 하는 저 물 리 목적 할 수 없습니다. 경 지. 하는 저 물 개 , 형 또는 가공할 수 없습니다. Doctoral ThesisYun Seop ShinSchool of Energy and Chemical Engineering(Energy Engineering)Ulsan National Institute of Science and Technology2023Phase Modulation of Emissive Materials forEfficient Blue Perovskite Light-Emitting DiodesYun Seop ShinSchool of Energy and Chemical Engineering(Energy Engineering)Ulsan National Institute of Science and TechnologyPhase Modulation of Emissive Materials forEfficient Blue Perovskite Light-Emitting DiodesiAbstractRecently, metal halide perovskites (MHPs) have drawn considerable research interest as next-generation materials for optoelectronic device applications in solar cells, lasing, and light-emitting diodes (LEDs). Especially, for lighting and future-generation displays, MHPs have emerged, takingadvantage of their excellent optoelectrical properties such as superior luminescence efficiency, narrow linewidth, easy band gap tunability, and defect tolerance. Tremendous developments have been made in enhancing the performance of perovskite LEDs (PeLEDs); external quantum efficiencies (EQEs) have exceeded 20% for green and red emissive PeLEDs. However, the lagged efficiency of blue emissive PeLEDs has hampered practical display applications.Blue emissive perovskite materials inevitably require the incorporation of chlorine that induces the enlarged optical bandgap entailing a relatively deeper highest occupied molecular orbital (HOMO)energy level and detrimental trap states within the bandgap, resulting in unbalanced energy level alignment with adjacent charge transporting layers in the device and poor photoluminescence quantum yield (PLQY) and stability of blue emissive perovskite materials, respectively. Therefore, the development of chlorine-less blue emissive perovskite materials is imperative for future full-color displays. Dimensional engineering is one effective approach, and quasi-two-dimensional (quasi-2D) perovskites can be an exceptional candidate. This thesis covers effective approaches for the modulation of 2D perovskite phases to realize efficient and stable blue PeLEDs.In chapter 2, an effective interfacial engineering strategy was introduced to guide the formation ofwell-grown 2D perovskite phases and minimize the detrimental chemical damage from the acidic poly(3,4-ethylenedioxythiophene):polystyrene sulfonic acid (PEDOT:PSS) substrate. Zwitterion additive, L-phenylalanine was incorporated into the PEDOT:PSS. This additive formed bidentate coordination with uncoordinated Pb2+ in the overlying perovskite layer which facilitates the growth of 2D perovskite phases and even thoroughly passivates the interfacial defect states. Moreover, attenuated acidity of the PEDOT:PSS, suppresses the chemical etching of indium tin oxide (ITO) substrates, reducing exciton quenching pathways. Finally, we realized efficient sky-blue emissive PeLEDs by blocking energy loss due to defect states through and designing an ideal energy landscape.In chapter 3, a multifunctional passivating molecule containing the electron-withdrawing chlorine atom and the phosphonic acid as a functional and anchoring group was introduced as an alternative to the PEDOT:PSS hole injection layer (HIL). It was observed that this molecule formed a self-assembled monolayer (SAM) on the ITO substrate, significantly increasing the work function (WF) of the ITO electrode. Strongly chemisorbed SAM inhibited the chemical etching of the ITO electrode, suppressingexciton quenching pathways. Besides, the incorporated chlorine atom induced an orbital coupling with the interfacial uncoordinated Pb2+ detect state, passivating them and thereby changing the optical iiproperties of the perovskite layer. Finally, due to these synergetic effects originating from a well-designed SAM molecule, we realized efficient pure-blue emissive PeLEDs.In chapter 4, the effect of the surface polarity of the underlying HIL on the crystallization dynamics of 2D perovskite phases was thoroughly studied, and optimized the surface polarity to realize the optimal substrate for efficient PeLEDs. It was confirmed that the additive, L-dopa involving hydroxyl groups coordinate chemical bonding with the sulfonate group of PSS moieties, leading to a higher surface polarity for the PEDOT:PSS substrate. On that substrate, the formation of higher-n 2D perovskite phases was thermodynamically unfavorable due to the smaller formation energy, and lower-n-dominated phase distribution induced a hypsochromically shift in the luminescence spectrum. Finally,we controlled the dual additives with L-phenylalanine which provide a well-matched electronic band structure, boosting the performance of pure-blue emissive PeLEDs.In chapter 5, a facile halide and phase modulating approach to design an ideal energy-transfer tunnel structure with flawless quasi-2D perovskites was introduced. This post-treatment entails the halide-exchange reaction with the assistance of haloalkane molecules and strong nucleophile molecules. The detached chlorides spontaneously exchanged with bromides in perovskites and further intrude the chlorine vacant sites, resulting in efficient PL and color stability. Furthermore, the spontaneous phase rearrangement occurred via merging between neighboring low-n 2D phases to higher-n 2D phases. It modulated the landscape of the energy-transfer funnel with the narrowed 2D phase distribution that can minimize the detrimental exciton losses. Finally, we realized efficient deep-blue emissive PeLEDs with all synergetic effects from the proposed halide and phase rearrange treatment.In chapter 6, a facile halide post-treatment was also introduced to realize efficient bulk blue emissive perovskite films. The halide compositions of bulk perovskite films were finely controlled for the desired emission colors. This spontaneous halide exchange process induced the recrystallization of rough perovskite surface, providing a fully covered and smooth perovskite film. Finally, we realized highly luminescent blue emissive bulk PeLEDs with long operating lifetimes under a high-level current injection without halide segregation.iiiivContentsAbstract ............................................................................................................................................ iContents .......................................................................................................................................... ivList of Figures ................................................................................................................................. viList of Tables................................................................................................................................. xviList of Abbreviations .................................................................................................................. xviiiCHAPTER 1. Introduction ............................................................................................................. 11.1. Structure and Properties of Metal Halide Perovskite ................................................ 11.1.1. Structure of Metal Halide Perovskite ................................................................... 11.1.2. Properties of Metal Halide Perovskite ................................................................. 21.2. Metal Halide Perovskite Light-Emitting Diodes ........................................................ 61.2.1. Operating Principles of Metal Halide Perovskite Light-Emitting Diodes.............. 61.2.2. Characteristics of Metal Halide Perovskite Light-Emitting Diodes....................... 71.3. Blue Emissive Metal Halide Perovskite Light-Emitting Diodes .............................. 101.3.1. Limitations of Blue Emissive Metal Halide Perovskite Materials ....................... 101.3.2. Structure and Properties of Quasi-Two-Dimensional Perovskite Materials ......... 13CHAPTER 2. Manipulated Interface for Enhanced Energy Cascade in Quasi-2D Blue Perovskite Light-Emitting Diodes ................................................................................................................... 182.1. Research Background............................................................................................... 182.2. Experimental Details ................................................................................................ 202.3. Results and Discussion.............................................................................................. 222.4. Conclusion................................................................................................................. 35CHAPTER 3. A Multifunctional Self-Assembled Monolayer for Highly Luminescent Pure-Blue Quasi-2D Perovskite Light-Emitting Diodes ................................................................................ 363.1. Research Background............................................................................................... 36v3.2. Experimental Details ................................................................................................ 383.3. Results and Discussion.............................................................................................. 403.4. Conclusion................................................................................................................. 55CHAPTER 4. Pure-Blue Electroluminescence of Quasi-2D Perovskites with Modulated Phase Distribution via Surface Polarity of PEDOT:PSS......................................................................... 564.1. Research Background............................................................................................... 564.2. Experimental Details ................................................................................................ 584.3. Results and Discussion.............................................................................................. 614.4. Conclusion................................................................................................................. 76CHAPTER 5. Phase Rearrangement for Minimal Exciton Loss in Quasi-2D Perovskite toward Efficient Deep-Blue LEDs via Halide Post-treatment................................................................... 775.1. Research Background............................................................................................... 775.2. Experimental Details ................................................................................................ 795.3. Results and Discussion.............................................................................................. 825.4. Conclusion................................................................................................................. 98CHAPTER 6. Highly Stable Bulk Perovskite for Blue LEDs with Anion-exchange Method...... 996.1. Research Background............................................................................................... 996.2. Experimental Details .............................................................................................. 1016.3. Results and Discussion............................................................................................ 1036.4. Conclusion................................................................................................................118CHAPTER 7. Summary ...............................................................................................................1197.1. Summary..................................................................................................................119References.................................................................................................................................... 121Acknowledgements ...................................................................................................................... 139viList of FiguresFigure 1.1. (a) Unit cell of the ABX3 perovskite crystal structure (A-site: monovalent cations, B-site: divalent cations, and X-site: halide anions) and (b) correlations between Goldschmidt tolerancefactor(t)andperovskiteGoldschmidt tolerance factor (t) for different A-site monovalent cations in perovskite system.Figure 1.3. (a) Electronic structure of the ideal APbX3 perovskite. (b) Schematic representation of the halide p-orbital and Pb 6s-orbital energy levels (left) and the correlation between halide p-orbital energy levels and bandgap (right).Figure 1.4. (a) Electronic structure of typical semiconductor (GaAs, CdTe) (left) compared to APbX3lead halide perovskite crystal (right). (b) Absolution PLQY as a function of the concentration of surface halide vacancy. (c) Absorption coefficient of lead halide perovskite thin film compared with typical semiconductors. (d) Calculated exciton binding energies of metal halide perovskites.Figure 1.5. FWHM of various light-emitting materials, organic emitters, inorganic QDs, and metal halide perovskite.Figure 1.6. (a) Schematic illustration of structure and operation of typical PeLEDs and (b) energy level diagram of various materials for HTL, ETL, and perovskite emitting layer.Figure 1.7. A typical structure of PeLEDs. The possible layers constituting a PeLED (left) and their contributions to various processes (right) are labeled.Figure 1.8. Development of state-of-the-art PeLEDs.Figure 1.9. (a) Scanning electron microscope (SEM) images of perovskite films with different chlorine ratios. (b) Electronic structure for CsPbCl3 with a surface chloride vacancy, CsPbBr3 with a surface bromide vacancy, and CsPbI3 with a surface iodide vacancy. (c) The evolution of EL spectra over operation time.Figure 1.10. (a) A typical band structure of blue emissive PeLEDs. (b) Schematic illustration of a hole transporting behavior between hole-injecting layer (HIL) and HTL. (c) The trend of turn-on voltage for some representative blue emissive PeLEDs.Figure 1.11. (a) Schematic crystal structures of quasi-2D perovskites with different n-values. (b) Ruddlesden-Popper (RP) 2D perovskite phase with the large-sized monoamine organic viication for A’-site and (c) Dion-Jacobson (DJ) 2D perovskite phase with the large-sized diamine organic cation for A’-site.Figure 1.12. Chemical structures of reported large-sized monoamine organic cations for RP 2D perovskite phase.Figure 1.13. Chemical structures of reported large-sized diamine organic cations for DJ 2D perovskite phase.Figure 1.14. (a) Electronic properties of quasi-2D perovskites, representing the quantum- and dielectric-confinement effects. (b) PL spectra of pure 2D phase with different n-values.Figure 1.15. (a) FRET theory with the distance (d) between the donor and acceptor. Absorption and PL spectra of donor (n = 1) and acceptor (n = 2) phases. (b) Schematic representation of energy transfer process within quasi-2D perovskite materials. (c) Transient absorption (TA) spectra of quasi-2D perovskite films.Figure 1.16. (a) Schematic illustration of the formation of exciton-polarons in 2D perovskites. (b) Arrhenius plot of experimental ln(kdis) as a function of negative reciprocal of temperature (–1/T) for n = 2 2D perovskites. (The red dashed line shows the simulated result by using steady-state Eb of 170 meV). (c) Absorption and PL spectra of pristine and phase rearranged quasi-2D perovskites.Figure 2.1. Optical and geometrical properties of quasi-2D perovskite films deposited on p- and m-PEDOT:PSS. (a) UV absorption spectra and (b) PL spectra. (excitation wavelength was 375 nm) (c) Schematic illustration of coordination bonding between L-phenylalanine and a quasi-2D perovskite. (d) XPS spectra corresponding to Pb 4f of 10 nm-thick perovskite films, and (e, f) GIWAXS patterns of 10 nm-thick perovskite films deposited on p- and m-PEDOT:PSS HILs.Figure 2.2. Optimization of PLQY of deposited quasi-2D perovskite films. (a) PL spectra and (b) PLQYs for perovskite films deposited on HILs with increasing L-phenylalanine concentration.Figure 2.3. (a) XPS spectra corresponding to Pb 4f peaks of PbBr2 and L-phenylalanine-treated PbBr2films. (b) FTIR spectra of L-phenylalanine and L-phenylalanine in presence of PbBr2.Figure 2.4. (a) Line-cut profiles for the GIWAXS patterns of quasi-2D perovskite films deposited on p- and m-PEDOTL:PSS HILs. (b) XRD patterns of quasi-2D perovskite films deposited on p- and m-PEDOT:PSS HILs.Figure 2.5. Influence of HIL acidity on the exciton quenching process. (a, b) PL kinetic profiles of viiiquasi-2D perovskite films with p- and m-PEDOT:PSS with different substrates on (a) bare glass and (b) ITO. (c) XPS spectra corresponding to In 3d for p- and m-PEDOT:PSS on ITO substrates. (d) Schematic illustrations of the exciton quenching process caused by diffused In species.Figure 2.6. Perovskite film stability. Photographs of blue emissive quasi-2D perovskite films deposited on p- and m-PEDOT:PSS using different substrates including bare glass (left) and ITO (right) under λ = 365 nm UV excitation.Figure 2.7. Ultrafast dynamics of photogenerated charge carriers in quasi-2D perovskite. TA spectra of quasi-2D perovskite films deposited on (a) p- and (b) m-PEDOT:PSS at variable time delays as noted in the panels. The corresponding normalized, steady-state UV-vis absorption (black, solid) and PL (black, dotted) spectra are also presented in each panel for comparison. Also included are magnified portions of the PL spectra (gray, dashed) obtained from independent measurements depicting the PL peak features attributed to n = 2 and n = 3 2D phases. (c) Kinetic profiles obtained from TA spectra at selected wavelengths representing the temporal behavior of bleached excitonic states corresponding to different phases. The profiles at 419 nm, 446 nm, 463 nm, and 492 nm correspond to the phases of n = 2, n = 3, bulk, and the defect states, respectively. The signal intensities have been scaled for comparison relative to the peak intensities of the n = 2 phase. (d) Schematic energy transfer mechanism in quasi-2D perovskites on m-PEDOT:PSS.Figure 2.8. Characteristics of PeLEDs. (a) Schematic flat-band energy diagram of the optimized PeLEDs, (b) J-V-L plot, (c) EQE-J curves, (d) EL spectra (inset shows a photograph of an operating PeLED) (e) Performance reproducibility and (f) J-V hysteresis of PeLEDs based on p- and m-PEDOT:PSS.Figure 2.9. UPS spectra of p- and m-PEDOT:PSS. (a) Secondary edge region and (b) valence band edge plotted relative to a Au reference.Figure 2.10. SCLC J-V curves of quasi-2D perovskite hole-only-diodes deposited on (a) p- and (b) m-PEDOT:PSS for the device architecture of ITO/p- and m-PEDOT:PSS/perovskite/MoO3/Ag.Figure 2.11. EL spectra at different injected current densities for PeLEDs using (a) p- and (b) m-PEDOT:PSS HILs. EL spectra at different time intervals for PeLEDs using (c) p- and (d) m-PEDOT:PSS HILs.Figure 2.12. Operational stability of PeLEDs using p- and m-PEDOT:PSS under a continuous constant current of 1.5 mA cm–2.ixFigure 3.1. Synthetic scheme for 36ClCzEPA.Figure 3.2. XPS spectra corresponding to (a) C 1s, (b) O 1s, (d) P 2p, and (d) Cl 2p of the 36ClCzEPA SAM on ITO substrates.Figure 3.3. (a) UPS spectra of the bare ITO substrates (black) and ITO/36ClCzEPA (blue). (b) Schematic energy level diagram of the ITO/36ClCzEPA. Electrostatic potential profiles by DFT-PBE for (c) bare ITO (111) and (d) ITO with a 36ClCzEPA SAM. The side view of each model structure is shown in the inset (In, pink; Sn, gray; O, red; C, brown; N, blue; Cl, green-yellow; P, violet; H, light pink).Figure 3.4. Optical properties of the perovskite films. (a) The normalized steady-state UV–vis absorption and (b) second derivative of the absorption spectra, (c) photoluminescence (PL) emission spectra of perovskite films with the PEDOT:PSS and 36ClCzEPA SAM as HIL. The inset shows a photograph of the corresponding perovskite films under λ = 365 nm UV excitation. DFT-PBE band structures of heterointerfaces of the layered 2D-CsPbBr3 with the (d) H- and (e) Cl-terminated SAMs. In the band structural plots, the contribution from the SAM molecule is shown by blue circles. Isosurfaces for the Bloch orbitals indicated by the arrow in each inset for the layered 2D-CsPbBr3 with the (f) H- and (g) Cl-terminated SAMs. The yellow and blue colors indicate positive and negative, respectively. (Cs, green; Pb, dark gray; Br, brown; O, red; C, dark brown; N, blue; Cl, green-yellow; P, violet; H, light pink).Figure 3.5. Photographs of water droplets in contact with (a) PEDOT:PSS and (b) 36ClCzEPA SAM.Figure 3.6. Side views of the optimized atomic structure of the layered 2D-CsPbBr3 surface with the (a) H- and (b) Cl-terminated SAM. The calculated binding energies are indicated. (Cs, green; Pb, dark gray; Br, brown; O, red; C, brown; N, blue; Cl, green-yellow; P, violet; H, light pink).Figure 3.7. Normalized steady-state PL emission spectra of 2D perovskite films deposited on the 36ClCzEPA and 2PACz SAMs.Figure 3.8. Geometrical properties of perovskite films. AFM images of (a) ITO/PEDOT:PSS, (b) ITO/36ClCzEPA SAM, and 2D perovskite films deposited on (c) ITO/PEDOT:PSS and (d) ITO/36ClCzEPA SAM (The scale bars are 1 µm). GIWAXS patterns of 2D perovskite films deposited on the (e) PEDOT:PSS and (f) 36ClCzEPA SAM HILs.Figure 3.9. Line-cut profiles for the GIWAXS patterns of 2D perovskite films deposited on the PEDOT:PSS and 36ClCzEPA SAM HILs.xFigure 3.10. Characteristics of trap densities and color stabilities. (a) PL kinetic profiles of perovskite films with the PEDOT:PSS and 36ClCzEPA SAM HILs on ITO substrates. (b, c) SCLC J–V curves of 2D perovskite hole-only-diodes deposited on the PEDOT:PSS and 36ClCzEPA SAM with the device structure of ITO/HIL/2D perovskite/MoO3/Ag. (d) XPS spectra displaying indium (In) 3d signals in the 2D perovskite film on the PEDOT:PSS and 36ClCzEPA. PL spectra of 2D perovskite films before and after (e) being kept under ambient conditions for 1 day and (f) thermal annealing at 100 °C for 20 min. The inset shows a photograph of the corresponding perovskite films under λ = 365 nm UV excitation.Figure 3.11. Device characteristics of PeLEDs. (a) Cross-sectional view for the PeLEDs device with the 36ClCzEPA SAM HIL. (b) EL spectra (inset shows photographs of operating PeLEDs), (c) current density–voltage–luminance, (d) current density–EQE curves, and (e) histogram of PeLEDs using the PEDOT:PSS and 36ClCzEPA SAM HILs. (f) Current density–EQE curves of PeLEDs with the PEDOT:PSS and 36ClCzEPA SAM HILs before and after the aging process. The aging process involved storing the fabricated PeLEDs under nitrogen for 1 day.Figure 3.12. Schematic illustrations of the hole-injection process in PeLEDs using the PEDOT:PSS and 36ClCzEPA SAM HILs.Figure 3.13. EL spectra at different injected current densities for PeLEDs using the (a) PEDOT:PSS and (b) 36ClCzEPA SAM HILs and (c) changes in the EL emission wavelength of PeLEDs as the current density increases. (d) Operational stability of PeLEDs using the PEDOT:PSS and 36ClCzEPA SAM HILs under a constant current density with an initial luminance of 100 cd m–2. The normalized EL spectra for PeLEDs using the (e) PEDOT:PSS and (f) 36ClCzEPA SAM HILs during the operational stability under a constant current density with an initial luminance of 100 cd m–2.Figure 4.1. (a) Chemical structures of additives, L-phenylalanine, L-tyrosine, and L-dopa. The contact angles of (b, c, d) water and (e, f, g) glycerol droplet on the PEDOT:PSS with different additives, (b, e) L-phenylalanine, (c, f) L-tyrosine, and (d, g) L-dopa.Figure 4.2. XPS spectra corresponding to (a) S 2p and (b) O 1s, (c) FTIR, and (d) zeta potential of PEDOT:PSS with different additives, L-phenylalanine, L-tyrosine, and L-dopa. (e) Schematic illustration of the chemical interaction between PSS moieties and additives.Figure 4.3. Optical properties of perovskite films. (a) Normalized, steady state UV–vis absorption and (b) second derivative of the absorption spectra. (c) Steady-state PL emission spectra (inset xishows a photograph of deposited perovskite films on the PEDOT:PSS with different additives). (d) Proportion of excitonic peaks (nx/n2) corresponding to each 2D phase calculated from the absorption spectra. (e) Normalized XRD pattern (XRD spectra have been normalized for n = 2 phase) of perovskite films deposited on the PEDOT:PSS with different additives, L-phenylalanine, L-tyrosine, and L-dopa. (f) Calculated formation energy of quasi-2D perovskites with different additives, L-phenylalanine and L-dopa corresponding to n = 2 and 4 phases.Figure 4.4. XPS spectra corresponding to (a) Pb 4f, (b) Br 3d, and (c) Cl 2p, and (d) the measured halide ratio of perovskite films deposited on the PEDOT:PSS with different additives, L-phenylalanine, L-tyrosine, and L-dopa.Figure 4.5. FWHM of the lower-n phase peak for perovskite films deposited on the PEDOT:PSS with different additives, (a) L-phenylalanine, (b) L-tyrosine, and (c) L-dopa.Figure 4.6. (a) DFT calculation model for the quasi-2D perovskite with (a) L-phenylalanine and (b) L-dopa to n = 2 phase. For the clear view, Cs, Pb, N, C, Br, Cl, O, and H atoms are colored purple, dark grey, blue, grey, brown, green, red, and white, respectively.Figure 4.7. (a) DFT calculation model for the quasi-2D perovskite with (a) L-phenylalanine and (b) L-dopa to n = 4 phase. For the clear view, Cs, Pb, N, C, Br, Cl, O, and H atoms are colored purple, dark grey, blue, grey, brown, green, red, and white, respectively.Figure 4.8. Calculated binding energy of quasi-2D perovskites with different additives, L-phenylalanine and L-dopa corresponding to n = 2 and 4 phase.Figure 4.9. Characteristics of PeLEDs. (a) Schematic flat-band energy diagram of the optimized PeLEDs, (b) EL spectra, (c) current density–voltage–luminance, and (d) current density–EQE curves for PeLEDs with PEDOT:PSS as HIL with different additives, L-phenylalanine, L-dopa, and mixed condition (L-phenylalanine:L-dopa = 80:20 v/v%). The inset shows a photograph of an operating PeLEDs. (e) Operating stability and (f) time-dependent evolution of EL spectrum under a continuous constant current density of 1 mA cm–2 for PeLEDs using the PEDOT:PSS with different additives, L-dopa (left) and mixed condition (right).Figure 4.10. Characteristics of PeLEDs. (a) Current density–voltage–luminance, (b) current density–EQE curves, and (c) EL spectra for PeLEDs with PEDOT:PSS as HIL with additive, L-tyrosine. The inset shows a photograph of an operating PeLEDs.Figure 4.11. UPS spectra of the PEDOT:PSS with different additives, L-phenylalanine, L-dopa, and mixed condition (L-phenylalanine:L-dopa = 80:20 v/v%): (a) Secondary electron cutoff xiiregion and (b) valance band edge plotted relative to a Au reference.Figure 4.12. Contact angle of a water droplet on the PEDOT:PSS with different volume percent of dual additives, L-phenylalanine and L-dopa (L-phenylalanine:L-dopa = x:100-x) and PL emission peaks of perovskite films deposited on the corresponding PEDOT:PSS.Figure 4.13. CIE coordinate of PeLEDs using the PEDOT:PSS with different additives, L-phenylalanine, L-tyrosine, L-dopa, and mixed condition (L-phenylalanine:L-dopa = 80:20 v/v%).Figure 5.1. Schematic illustration of halide and phase rearrange treatment process to obtain deep-blue color quasi-2D perovskite films.Figure 5.2. Optical properties of deposited control and HPR-treated perovskite films. (a, b) AFM images, (c) Steady-state PL emission spectra and corresponding PLQYs. Excitation wavelength was 375 nm. (d) XPS spectra corresponding to Pb 4f. (e) Schematic electronic levels. (f) PL spectral stabilities of the perovskite films deposited on glass/PEDOT:PSS. Presented are PL spectral evolution in air for 2 hours with an RH up to 50% (left) and during thermal annealing at 120 °C (right).Figure 5.3. AFM image of the as-prepared sky-blue color quasi-2D perovskite film.Figure 5.4. (a) Optical bandgaps and (b, c) UPS spectra of deposited control and HPR-treated perovskite films. (b) Secondary edge region and (c) valance band edge plotted relative to a Au reference.Figure 5.5. Rearrangement of phase distribution in quasi-2D perovskites with the HPR treatment. (a, b) UV–vis absorption spectra of deposited, control and HPR-treated perovskite films. The second derivatives of the absorption spectra (dotted) are also presented. (c, d) GIWAXS patterns of deposited, control and HPR-treated perovskite films.Figure 5.6. (a) Line-cut profiles for the GIWAXS patterns and (b) XRD patterns of deposited control and HPR-treated perovskite films.Figure 5.7. XPS spectra corresponding to the N 1s peak for control and HPR-treated perovskite films. The intensities are normalized to corresponding Pb 4f peaks.Figure 5.8. Ultrafast dynamics of photogenerated carriers in quasi-2D perovskite. TA spectra of the control (a) and HPR-treated (b) perovskite films at representative time delays. The spectra are normalized to the DA value of the n = 2 phase. Time delays are given in the panels. The corresponding UV–vis absorption (solid) and PL (dashed) spectra are also presented in each upper panel for comparison. (c) Kinetic profiles obtained from TA spectra with fit xiiilines on an early timescale at representative wavelengths to the various phases. The profiles are normalized to the DA value of the n = 2 phase. (d) Intensity profiles obtained from TCSPC with fit lines for the control (red open circle) and HPR-treated (blue open circle) perovskite films upon excitation at 375 nm. (e) Schematic diagram of exciton/carrier dynamics in quasi-2D perovskites.Figure 5.9. Kinetic profiles obtained from TA spectra with fit lines for the entire time window for the control and HPR-treated perovskite films at selected wavelengths assigned to the various phases. The profiles have been normalized for n = 2 phase.Figure 5.10. Kinetic profiles obtained from independent measurements with improved S/N ratio on very early timescale for the HPR-treated perovskite film at selected wavelengths assigned to the n = 3 (443 nm) and n ≥ 4 phases (461 nm).Figure 5.11. Characteristics of PeLEDs. (a) Schematic flat-band energy diagram of the optimized PeLEDs, (b) EL spectra. The inset shows a photograph of an operating PeLED. (c) CIE coordinates. (d) J–V–L plot. (e) EQE–J curves. (f) Performance reproducibility.Figure 5.12. SCLC J–V curves of control and HPR-treated perovskite hole-only-diodes with the structure of ITO/PEDOT:PSS/control and HPR-treated perovskite/MoO3/Ag.Figure 5.13. (a, b) Normalized EL spectra at different injected current densities. (c) Operating stability under a continuous constant voltage of 4.0 V.Figure 6.1. Scheme and optical properties. (a). Scheme of the anion exchange method for blue-emitting bulk perovskite films. (b). UV–vis absorption and PL spectra of reference (Ref. 520) and anion-exchanged bulk perovskite films (A.E. 470 and A.E. 490 conditions). (c). Thermal stability measurements; PL spectra evolution of deposited CsPbBr3, MAPbBr3, and FAPbBr3 perovskite films without and with thermal annealing at 100 °C.Figure 6.2. Photographs of anion-exchanged films with the assistance of TOP, THP, and TBP at the same reaction condition (reaction time and concentration).Figure 6.3. PL spectra of reference (Ref. 520) and anion-exchanged bulk perovskite films.Figure 6.4. (a) Tauc plot of reference (Ref. 520) and anion-exchanged bulk perovskite films. Measured optical bandgaps are indicated. (b) TRPL lifetime at detection wavelengths of 520, 490, and 470 nm for reference (Ref. 520) and anion-exchanged bulk perovskite films deposited on glass/PEDOT:PSS substrates. (c) PLQY of reference (Ref. 520) and anion-exchanged bulk perovskite films.Figure 6.5. Photographs of CsBr and CsCl solution dissolved in DMSO. The concentration of CsX xivsolutions was 30 mg mL–1, respectively (left). Photographs of CsPb(TFA/Br)3 and CsPb(TFA/Cl)3 solution dissolved in DMSO. The concentrations of solutions were 0.2 M respectively (right).Figure 6.6. XPS spectra corresponding to (a). Br 3d, and (b). Cl 2p, and (c). halide ratio in reference (Ref. 520) and anion-exchanged bulk perovskite films based on XPS measurement. Reconstructed TOF-SIMS elemental 3D maps for Cl– halides traced in the depth profile;(d). reference (Ref. 520) and anion-exchanged bulk perovskite films, (e). A.E. 490 and (f). A.E. 470 conditions. The image size is 100 μm × 100 μm in the x-y plane.Figure 6.7. XPS spectra corresponding to P 2p of reference (Ref. 520) anion-exchanged bulk perovskite films.Figure 6.8. Energy-dispersive X-ray (EDX) spectroscopy images corresponding elemental mapping of Cs, Pb, Br, and Cl in (a-d) reference (Ref. 520), (e-h) anion-exchanged bulk perovskite films, A.E. 490 condition and (i-l) anion-exchanged bulk perovskite films, A.E. 470 condition.Figure 6.9. Negative ion TOF-SIMS spectra and depth profiles of (a) reference (Ref. 520) and anion-exchanged bulk perovskite films, (b) A.E. 490 and (c) A.E. 470 conditions.Figure 6.10. Spectral stability measurements; PL spectra evolution of (a) reference (Ref. 520) and anion-exchanged bulk perovskite, (b) A.E. 490 and (c) A.E. 470 conditions after continuous thermal annealing at 100 °C for different times.Figure 6.11. XRD pattern of reference (Ref. 520) and anion-exchanged bulk perovskite films.Figure 6.12. SEM surface images of (a) reference and anion-exchanged bulk perovskite films, (b) A.E. 490 and (c) A.E. 470 conditions. AFM surface images of (d) reference (Ref. 520) and anion-exchanged bulk perovskite films, (e) A.E. 490 and (f) A.E. 470 conditions.Figure 6.13. SEM surface images of (a) reference and dipped films into (b) chloroform and (c) TBP for 5 minutes.Figure 6.14. PL spectra and photographs of reference and dipped films into chloroform and TBP for 5 minutes.Figure 6.15. Cross-sectional SEM images of (a) reference (Ref. 520) and anion-exchanged bulk perovskite films, (b) A.E. 490 and (c) A.E. 470 conditions.Figure 6.16. AFM images of PEDOT:PSS layer (a) w/o and (b) w/ dipping into TBP mixed CF solution.Figure 6.17. PeLEDs characteristics: (a) Band energy diagram of a device, (b) EL spectra, (c) J–V–Lxvplot, (d) CE–V, (e) EQE–V, and CIE color space coordinates of PeLEDs based on reference (Ref. 520) and anion-exchanged films, A.E. 490 and A.E. 470 conditions.Figure 6.18. UPS spectra (left: secondary electron cutoff region and right: onset region) of reference (Ref. 520) and anion-exchanged bulk perovskite films, A.E. 490 and A.E. 470 conditions.Figure 6.19. Performance uniformity of PeLEDs with anion-exchanged bulk perovskite films, A.E. 490 condition.Figure 6.20. EL spectra operated under different applied voltages of the optimized PeLEDs with anion-exchanged polycrystalline perovskite films (a) A.E. 490, (b) A.E. 470 conditions, and (c)without TFA precursors.Figure 6.21. (a) Lifetime (T50) and (b) CIE coordinates of anion-exchanged polycrystalline perovskite LED devices at a constant current density of 20 mA cm–2.xviList of TablesTable 2.1. Triexponential fitting parameters for PL lifetimes of quasi-2D perovskite films with p- and m-PEDOT:PSS on bare glass and ITO substrates.Table 2.2. pH values of p- and m-PEDOT:PSS solutions.Table 2.3. Multi-exponential fit parameters for the kinetic profiles representative of various phases for perovskite films deposited on pristine and modified PEDOT:PSS substrates.Table 2.4. Summary of device characteristics of optimized PeLEDs using p- and m-PEDOT:PSS HILs.Table 3.1. Triexponential fitting parameters for PL lifetimes of perovskite film deposited on the PEDOT:PSS and 36ClCzEPA SAM on ITO substrates.Table 3.2. Summary of device characteristics of optimized PeLEDs using the PEDOT:PSS and 36ClCzEPA SAM HILs before and after the aging process.Table 4.1. Summarized contact angles of water and glycerol droplets and calculated solid/liquid interfacial free energies on the PEDOT:PSS with different additives; L-phenylalanine, L-tyrosine, and L-dopa. \\\\x00\\\\x00\\\\x00 represent the solid/liquid interfacial free energy.Table 4.2. Summarized device performances of optimized PeLEDs with PEDOT:PSS as HIL with different additives, L-phenylalanine, L-dopa, and mixed condition (L-phenylalanine:L-dopa = 80:20 v/v%).Table 4.3. Summarized CIE coordinate of PeLEDs using the PEDOT:PSS with different additives, L-phenylalanine, L-tyrosine, L-dopa, and mixed condition (L-phenylalanine:L-dopa = 80:20 v/v%).Table 5.1. Multi-exponential fit parameters for the kinetic profiles obtained from TA, representative of various phases for the control and HPR-treated perovskite films.Table 5.2. The multi-exponential fit parameters obtained for the kinetic profiles obtained from independent measurements with improved S/N ratio on very early timescale for the HPR-treated perovskite film at selected wavelengths assigned to the n = 3 (443 nm) and n ≥ 4 phases (461 nm).Table 5.3. Multi-exponential fit parameters for the intensity profiles obtained from TCSPC for the control and HPR-treated perovskite films.Table 5.4. Device characteristics of optimized PeLEDs using the control and HPR-treated perovskite xviifilms.Table 6.1. Summarized biexponential fitting parameters for TRPL lifetime of reference (Ref. 520) and anion-exchanged bulk perovskite films, A.E. 490 and A.E. 470 conditions, deposited on glass/PEDOT:PSS substrates.Table 6.2. Summarized energy level values for reference (Ref. 520) and anion-exchanged bulk perovskite films, A.E. 490 and A.E. 470 conditions, obtained from UPS spectra and Tauc plot.Table 6.3. Summarized device performances of optimized PeLEDs with reference and anion-exchanged bulk perovskite, A.E. 490 and A.E. 470 conditions, as emitting layers.xviiiList of AbbreviationsAbbreviation DescriptionMHP Metal halide perovskitePLQY Photoluminescence quantum yieldPCE Power conversion efficiencyFWHM Full width at half-maximumLED Light-emitting diodePeLED Perovskite light-emitting diodeHTL Hole-transporting layerETL Electron-transporting layerITO Indium tin oxideFTO Fluorine doped tin oxideEQE External quantum efficiencyIQE Internal quantum efficiencyCE Current efficiencyPE Power efficiencyHOMO Highest occupied molecular orbitalPL PhotoluminescenceEL ElectroluminescenceHIL Hole-injecting layerQuasi-2D Quasi-two-dimensionalRP Ruddlesden-PopperDJ Dion-JacobsonxixQW Quantum-wellFRET Förster resonance energy transferTA Transient absorptionPEDOT:PSS Poly(3,4-ethylenedioxythiophene):polystyrene sulfonic acidTPBi 1,3,5-tris(N-phenylbenzimidazol-2-yl) benzeneSEM Scanning electron microscopeAFM Atomic force microscopeXRD X-ray diffractionXPS X-ray photoelectron microscopyUPS Ultraviolet photoelectron microscopyTRPL Time-resolved PLGIWAXS Grazing-incidence wide-angle X-ray scatteringFTIR Fourier transform infraredTOF-SIMS Time of flight-secondary ion mass spectrometryEDX Energy-dispersive X-ray spectroscopyPIB Photoinduced bleachingIRF Instrumental response functionSCLC Space-charge-limited currentDFT Density functional theorySAM Self-assembled monolayer2PACz [2-(9H-carbazol-9-yl)ethyl]phosphonic acid36ClCzEPA [2-(3,6-dichloro-9H-carbazol-9-yl)ethyl]phosphonic acidWF Work functionVASP Vienna Ab initio simulation packagexxGGA Generalized gradient approximationPBE Perdew-Burke-ErnzerhofRMS Root-mean-squareCF ChloroformTBP TributylphosphineTHP TrihexylphosphineTOP TrioctylphosphineHPR Halide and phase rearrangeTFA TrifluoroacetateNTSC National Television System Committee1CHAPTER 1. Introduction1.1. Structure and Properties of Metal Halide Perovskite1.1.1. Structure of Metal Halide PerovskitePerovskite is a naturally forming mineral of calcium titanate with a chemical formula of CaTiO3, was named after Russian mineralogist, Lev Perovski. The structure of perovskite used in photovoltaic fields has a similar structure to CaTiO3. It was first successfully used in solid-state solar cells in 2012, since then most solar cells have used the desired combination of perovskite materials. As shown in Figure 1.1a, perovskite has a nearly cubic structure with the usual formula ABX3 which is composed of the A cation and the corner-sharing BX6 octahedra. Commonly, the A-site is monovalent cations such as methylammonium (MA+), formamidinium (FA+), and Cs+, the B-site is divalent metal cations such as Pb2+ and Sn2+, and the X-site is halide ions such as Cl–, Br–, and I–.Figure 1.1. (a) Unit cell of the ABX3 perovskite crystal structure (A-site: monovalent cations, B-site: divalent cations, and X-site: halide anions) and (b) correlations between Goldschmidt tolerance factor (t) and per2Perovskite crystal structure depends on the ionic radius and interaction of each component in ABX3. Goldschmidt tolerance factor (t) is a reliable geometric ratio to predict which structure is favorably formed. The Goldschmidt tolerance factor can be calculated from the ionic radii of each component using the following equation:\\\\x00 = \\\\x00\\\\x00\\\\x00 \\\\x00\\\\x00√\\\\x00 (\\\\x00\\\\x00\\\\x00 \\\\x00\\\\x00) (1-1)where \\\\x00\\\\x00, \\\\x00\\\\x00, and \\\\x00\\\\x00 are the radii of the A-site monovalent cations, the B-site divalent cations, and the X-site halide anions, respectively.The acceptable range of tolerance factor is 0.78 < t < 1.05 for the formation of the stable perovskite structure (Figure 1.1b). Above all, to have an ideal cubic structure, the tolerance factor should be included in the range of 0.90 < t < 1.00. Therefore, MA+, FA+, and Cs+ cations have been commonly used to obtain the stabilized perovskite structure as shown in Figure 1.2. A tolerance factor of perovskite structure can be determined with different radii of A-site cations. With ethylammonium (EA+) and ethylenediamine (EDA) cations are too large to form a cubic structure with tolerance factor t > 1, and tolerance factors with too small-sized cations such as Na+, K+, and NH4+ also fall outside the range for the stable perovskite structure (t < 0.8). They are prone to form a non-perovskite structure; orthorhombic or hexagonal structure. However, the mixed composition of A-site cations has been developed as an additive to stabilize the perovskite phase.Goldschm1.1.2. Properties of Metal Halide Perovskite3Metal halide perovskite (MHP) can have various combinations with varying compositions, as alluded to above, enabling easy bandgap tunability. This characteristic is mainly advantageous for diverse optoelectronic device applications. The electronic band structures of MHPs and their bandgap are defined by the inorganic metal-halide framework, BX6. The valance band comprises the s-lone pair of Pb2+ and halide p-orbitals, and the conduction band comprises Pb 6p and halide p-orbitals (Figure 1.3a). With Pb2+ as a B-site cation, the bandgap of MHP is altered with varying compositions of X-site halide anions due to the different halide p-orbitals (Figure 1.3b). Besides, the tilted inorganic metal-halide framework by steric and Coulombic interactions with small or large-sized A-site cations also represents the changed electronic band structure.Due to this electronic structure, MHP shows a strong tolerance to crystal defects, compared to other typical semiconductors (GaAs, CdTe) (Figure 1.4a). The ionic nature of perovskite is vulnerable to inducing defect states in the perovskite lattice, mainly through the formation of halide vacancies. Because intrinsic defects in MHP such as halide vacancies produce defect energy levels near or within the band edges, MHP retains its optical properties despite the presence of defect states. As shown in Figure 1.4b, surface chloride vacancies are observed to have a moderate effect on the photoluminescence quantum yield (PLQY) of perovskite, even so, perovskite exhibits a relatively strong tolerance to other halide vacancies with maintained their PLQY.Figure 1.3. (Moreover, MHP has a high absorption coefficient (Figure 1.4c), compared to other typical semiconductors, providing a chance to absorb sufficient light for the generation of excitons. Considering the unusual sharp slope of the curve near the bandgap of MHP, it also shows a relatively low Urbach energy which represents the absorption tail states. Besides, as shown in Figure 1.4d, MHP has a relatively low exciton binding energy, in particular, I–-based perovskite has very low exciton binding 4energy which is comparable to the thermal energy of room temperature (26 meV). These mentioned properties are advantageous for solar cell applications; a large number of excitons which are generated by sufficient light absorption, are easily dissociated into free charge carriers, recently achieving 25% of power conversion efficiency (PCE).Figure 1.4. (a) Electronic structure of typical semiconductor (GaAs, CdTe) (left) compared to APbX3 lead halide perovskite crystal (right). Reprinted with permission.5 Copyright 2020, Royal Society of Chemistry. (b) Absolution PLQY as a function of the concentration of surface halide vacancy. Reprinted with permission5MHP also have used as light-emitting material. Especially, it shows size-insensitively very narrow spectra with full width at half-maximum (FWHM) of approximately ~20 nm, compare to other light-emitting materials (Figure 1.5), and this highlights that MHP can be promising emitters in display fields. Synergistic advantages of MHP such as defect tolerance and easy bandgap tunability, as alluded to above, have been made in enhancing the performance of MHP-based light-emitting diodes (LEDs), achieving by over 20% for green and red emission.Figure 1.5. FWHM of various light-emitting materials, organic emitters, inorganic QDs, and metal halide perovskite. Reprinted with permission.9 Copyright 2016 National Academy of Sciences.61.2. Metal Halide Perovskite Light-Emitting Diodes1.2.1. Operating Principles of Metal Halide Perovskite Light-Emitting DiodesFigure 1.6. (a) Schematic illustration of structure and operation of typical PeLEDs and (b) energy level diagram of various materials for HTL, ETL, and perovskite emitting layer. Reprinted with permission.9 Copyright 2016 National Academy of Sciences.A LED is a device that emits light when current flows through the cathode and anode. Typically, conventional perovskite LEDs (PeLEDs) are comprised of anode, hole-transporting layer (HTL), perovskite emitting layer, electron-transporting layer (ETL), and cathode (Figure 1.6a). Indium tin oxide (ITO) and fluorine doped tin oxide (FTO) have been used as anodes and metals electrode such as Al, Ag, and Au have been used as cathodes, consisting suitable combinations according to the architecture of PeLEDs and the energy level of perovskite emitting materials. 4 steps are mainly occurred in PeLEDs, (1) charge carrier injecting, (2) charge carrier transporting, (3) charge carrier recombination, and (4) light emission. Charge carriers, electrons and holes are injected through the cathode and anode, respectively, and flow into the perovskite emitting layer along the ETL and HTL. These two interlayers, ETL and HTL, can facilitate the flow of charge carriers into the perovskite emitting layer by reducing potential barriers, and even block the leakage of charge carriers crossing the perovskite emitting layer. Therefore, the use of appropriate interlayers having well-matched energy levels with adjacent perovskite emitting layer and electrode is essential to develop the efficient and stable PeLEDs (Figure 1.6b). Especially, in the case of underlying HTL, HTL having different surface characteristics, directly affect the geometrical and optical properties of overlying perovskite emitting layer.71.2.2. Characteristics of Metal Halide Perovskite Light-Emitting DiodesThe crucial parameters that determine the performance of PeLEDs are the luminance (L), the external quantum efficiency (EQE), the current efficiency (CE), the power efficiency (PE), and the turn-on voltage (Von).Figure 1.7. A typical structure of PeLEDs. The possible layers constituting a PeLED (left) and their contributions to various processes (right) are labeled. Reprinted with permission.10 Copyright 2022 Nature Portfolio.Luminance (L)The luminance (L) describes the luminous intensity per unit area in a given direction (unit: cd m–2). On the assumption that condition is monochromatic light, the L is expressed as the following equation:\\\\x00 =\\\\x00\\\\x00= \\\\x00\\\\x00\\\\x00⋅ ∫ \\\\x00(\\\\x00)Φ\\\\x00,\\\\x00(\\\\x00)\\\\x00\\\\x00\\\\x00\\\\x00 (1-2)where the Φ, \\\\x00\\\\x00, \\\\x00(\\\\x00), and Φ\\\\x00,\\\\x00(\\\\x00) describe the luminous flux, maximum luminosity factor (683 lm W–1), the luminosity function at wavelength (representing the average spectral sensitivity of human visual perception of light), and the spectral radiant flux, respectively.8External quantum efficiency (EQE)The EQE describes the outcoupled photons per injected charge and is thus a key parameter for a LED device. There are two key factors that determine the EQE, the photon generation efficiency and the light extraction efficiency of generated photons (\\\\x00\\\\x00\\\\x00\\\\x00) into free space. The photon generation efficiency is known as the internal quantum efficiency (IQE), being the ratio of the number of photons generated in the perovskite emitting layer per unit time to the number of electrons injected into the device per unit time. The IQE, EQE, and \\\\x00\\\\x00\\\\x00\\\\x00 are expressed as the following equations:\\\\x00\\\\x00\\\\x00 = \\\\x00\\\\x00\\\\x00\\\\x00 × \\\\x00\\\\x00\\\\x00\\\\x00 (1-3)\\\\x00\\\\x00\\\\x00 = \\\\x00\\\\x00\\\\x00\\\\x00 × \\\\x00\\\\x00\\\\x00\\\\x00 × \\\\x00\\\\x00\\\\x00\\\\x00 = \\\\x00\\\\x00\\\\x00 × \\\\x00\\\\x00\\\\x00\\\\x00 (1-4)\\\\x00\\\\x00\\\\x00\\\\x00 = 1/(2\\\\x00\\\\x00) (1-5)where \\\\x00\\\\x00\\\\x00\\\\x00, \\\\x00\\\\x00\\\\x00\\\\x00, and \\\\x00 describe the number of injected charge carriers into the perovskite emitting layer, the ratio of the radiatively recombined excitons to the totally generated excitons into the device, and the refractive index of a layer that the generated photons will pass, respectively.Current efficiency (CE)The CE is another parameter that determines the efficiency of the LED, describing the ratio of the luminance (L) to the current density (J) of the LED (unit: cd A–1) as the following equation:\\\\x00\\\\x00 = \\\\x00/\\\\x00 (1-6)Power efficiency (PE)The PE describes the efficiency of the LED to convert electrical power into light. Also, it defines as the ratio of the input electrical power (\\\\x00\\\\x00\\\\x00 × \\\\x00) to the output optical power (\\\\x00\\\\x00\\\\x00\\\\x00) of the LED (unit: lm W–1)as the following equation:\\\\x00\\\\x00 =\\\\x00\\\\x00\\\\x00\\\\x00\\\\x00\\\\x00\\\\x00 × \\\\x00=\\\\x00\\\\x00\\\\x00\\\\x00ℏ\\\\x00\\\\x00\\\\x00 (1-7)9where ℏ\\\\x00 and \\\\x00 are the mean energy of the emitted photon and elementary charge, respectively.Turn-on voltage (Von)As the applied voltage (forward bias) increases, the charge carriers start to inject into the perovskite emitting layer through the cathode and anode, generating the photons. Then, the turn-on voltagedescribes the applied voltage that shows minimum detectable luminance intensity of approximately ~0.1 cd m–2. The turn-on voltage can be affected and facilitated by the charge-transporting behavior ofHTL and ETL, ensuring the reduction of the turn-on voltage. Besides, in the case of PeLEDs, the color of the perovskite emitting layer which is determined by its bandgap even influences the turn-on voltage. Blue color emissive perovskite showing a wide bandgap has a relatively deeper highest occupied molecular orbital (HOMO) energy level, hampering the flow of holes. Hence, it provides the highest turn-on voltage in PeLEDs, compared to other colors, green and red, limiting the efficiency of blue emissive PeLEDs.101.3. Blue Emissive Metal Halide Perovskite Light-Emitting Diodes1.3.1. Limitations of Blue Emissive Metal Halide Perovskite MaterialsMHPs have emerged as promising emitters in display fields, taking advantage of their fascinating properties such as defect tolerance, easy bandgap tunability, and narrow FWHM, as alluded to above. Substantial approaches have been made in enhancing the performance of PeLEDs in the past few years; EQEs have exceeded 20% for green and red emissive PeLEDs. However, the efficiency of blue emissive PeLEDs lags behind their counterparts (Figure 1.8). Historically blue electroluminescence (EL) remains challenging in all previous LED technologies and, practically, blue spontaneous light-emitting LEDs still fall short of commercialization.The incorporation of chlorine is indispensable for tuning the blue color perovskite materials. As shown in Figure 1.3b, the chlorine induces a wide bandgap of perovskite materials, resulting in the generation of photons having higher energy. However, from the incorporation of chlorine, some inherent drawbacks of blue emissive perovskite materials including color instability, poor PLQY, and energy level mismatching are the main challenges, hence, the development of new types of blue emissive perovskite materials is an essential part for future display.Figure 1.8. Development of state-of-the-art PeLEDs. Reprinted with permission.10 Copyright 2022 Nature Portfolio.11To provide the chlorine contents in the perovskite materials, lead chloride (PbCl2) precursor was commonly used. However, it shows poor solubility compared to other lead halide precursors (lead iodide, PbI2 and lead bromide, PbBr2), limiting the realization of deep-blue color emissive PeLEDs with a three-dimensional (3D) bulk type of perovskite. During the solution-processed deposition, the crystallization of perovskites becomes uncontrollable, obtaining rough and not fully-covered perovskite films (Figure 1.9a). Yoon et al.13 have reported a facile halide-exchange method to realize efficient sky- and deep-blue emissive bulk perovskite films. The spontaneous recrystallization process accompanied by a halide-exchange process has induced smooth and compact perovskite films, and further fabricated highly stable and luminescent blue PeLEDs. However, the large size of the perovskite grain is disadvantageous for suppressing bimolecular annihilation of the excitons, resulting in the poor efficiency of PeLEDs.Figure 1.9. (a) Scanning electron microscope (SEM) images of perovskite films with different chlorine ratios. Re12The chlorine vacancies are prone to be easily induced in the mixed halide (Br/Cl) perovskite materials, unfortunately, they inherently induce relatively deeper defect-state energy levels within the optical bandgap as shown in Figure 1.9b, significantly damaging the optical properties of perovskite materials despite the character of defect tolerance (Figure 1.4b). Besides, these halide vacancies appear to directly cause the instability of perovskite films. They activate halide migration serving as a hopping center and thereby inducing the color instability and poor PLQY of perovskite film. Furthermore, under the operation of the PeLEDs, they can irreversibly capture the injected charge carriers and suppress the radiative recombination in the perovskite emitting layer. Likewise, the facilitated halide migration degrades the device with color instability and makes the perovskite more vulnerable to external stimuli (Figure 1.9c).The chlorine-incorporated perovskite materials have a wide bandgap, representing a deep HOMO energy level (> 6.2 eV) (Figure 1.10a). The design of an optimal device structure with some possible HTLs having mismatched HOMO (5.2 eV < HOMO < 5.6 eV) energy levels has been difficult. As shown in Figure 1.10b, due to the large potential barrier from HTL to the blue emissive perovskite layer, the injected holes can be accumulated without flowing into the perovskite emitting layer. The unbalanced injection of charge carriers undermines the efficiency of PeLEDs through the non-radiative recombination losses. Then, inevitably, the turn-on voltage will be increased with the blue emissive perovskite materials having an enlarged bandgap (Figure 1.10c).Figure 1.10. (a) A typical band structure of blue emissive PeLEDs. Reprinted with permission.14 Copyright 20131.3.2. Structure and Properties of Quasi-Two-Dimensional Perovskite MaterialsThe chlorine-incorporated perovskite materials suffer from severe intrinsic chemical instability, remaining the critical bottleneck that hampers their further development for efficient blue emissive PeLEDs. To overcome this, quasi-two-dimensional (quasi-2D) perovskites have emerged as an alternative potential blue emissive material.Composition and structure of quasi-2D perovskiteQuasi-2D perovskites (Figure 1.11a) possess the chemical formula A’2An-1BnX3n+1 (n ≥ 1), where A’ is a large-sized monoamine organic cation (Figure 1.11b and Figure 1.12) (A’ is large-sized diamine organic cation, referring as Dion-Jacobson phase quasi-2D perovskite as shown in Figure 1.11c and Figure 1.13), A is a small-sized monovalent cation, B is a divalent cation, X is halide anions, and n refers the number of octahedral [BX6]4– layers. Inorganic octahedral [BX6]4– layers are sandwiched by large-sized organic cations, posing self-assembled multiple-quantum-well structures; inorganic octahedral layers act as quantum wells, while the layer consisting of large-sized organic cations act as barriers.Figure 1.11. (a) Schematic crystal structures of quasi-2D perovskites with different n-values. Reprinted with permission.17 Copyright 2021 Nature Portfolio. (b) Ruddlesden-Popper (RP) 2D perovskite phase with the large-sized monoamine organic cation for A’-site and (c) Dion-Jacobson (DJ) 2D perovskite phase with the large-sized diamine organic cation for A’-site. Reprinted with permission.18 Copyright 2021 Elsevier.14Figure 1.12. Chemical structures of reported large-sized monoamine organic cations for RP 2D perovskite phase. Reprinted with permission.19 Copyright 2020 Frontiers Media S.A.Figure 1.13. Chemical structures of reported large-sized diamine organic cations for DJ 2D perovskite phase. Reprinted with permission.20 Copyright 2021 Walter de Gruyter / Science Wise Publishing.15Photophysical properties of quasi-2D perovskiteFrom the “quantum-well” (QW) structures of quasi-2D perovskite, the quantum- and dielectric-confinement effects are strengthened. The difference in dielectric constant between the inorganic octahedral well and the barrier induces the dielectric confinement effect, directly affecting its exciton binding energy. The movement of the charge carriers is limited, and they will be confined within the inorganic octahedral well due to the compression of the carrier wave function in one direction, resulting in the increase of exciton binding energy (Eb) and bandgap of quasi-2D perovskite. Hence, in the case of n = 1, the dielectric confinement dominates with larger exciton binding energy and lower bandgap (Figure 1.14a), and the dielectric confinement becomes less dominant as the number of octahedral layers increases. Besides, easy bandgap tunability is possible by controlling the number of octahedral layers (Figure 1.14b). Thus, deep-blue color emissive PeLEDs can be realized by using less chlorine, minimizing the detrimental damages from the chlorine vacancies. Moreover, the geometrical properties of QW structures accompanied by large exciton binding energy complement the inherent characteristics of MHP such as low exciton binding energy and high charge carrier mobility. These features make quasi-2D perovskite materials to be appropriate for PeLED applications.Figure 1.14. (a) Electronic properties of quasi-2D perovskites, representing the quantum- and dielectric-confinement effects. Reprinted with permission.17 Copyright 2021 Nature Portfolio. (b) PL spectra of pure 2D phase with different n-values. Reprinted with permission.21 Copyright 2018 Nature Portfolio.16Optical properties of quasi-2D perovskiteEfficient energy transfer as well as effective exciton recombination is the striking characteristic ofquasi-2D perovskite, which contribute to the excellent optical properties. The short distance between the neighboring 2D phases facilitates sequential Förster resonance energy transfer (FRET), where energy is transferred from a donor with a higher bandgap to an acceptor with a lower bandgap (Figure 1.15a). This sequential FRET depends on the spectra overlap integral and electronic coupling, influenced by donor-acceptor distance. Therefore, the constraint of the spectral overlap between donor and acceptor phases results in the sequential energy transfer between the 2D phases of neighboring numbers of the octahedral layers; from n = 1 to n = 2 and from n = 2 to n = 3, as shown in Figures1.15b,c. In detail, photogenerated excitons undergo cascading energy transfer to the lowest-bandgap phases along with an energy-funneling pathway which is established by a graded band structure, resulting in efficient light emission of the desired color. This feature provides a chance to suppress thenonradiative recombination loss and obtain high film PLQYs even at low excitation intensity.This cascading energy transfer significantly contributes to the outstanding optical properties of quasi-2D perovskite materials. However, it is strongly affected by the distribution of 2D phases in the perovskite films, in particular, the formation of the lower-n 2D phases (i.e., n = 1 and 2) that would aggravate the optical properties of quasi-2D perovskites due to the strong exciton-phonon coupling and inefficient internal energy transfer, should be suppressed. As shown in Figure 1.16a, the layered-2D perovskite, especially the lower-n 2D phase is prone to form exciton-polarons then, the process of exciton dissociation into free carriers is promoted with significantly reduced exciton binding energy (Figure 1.16b). Therefore, the generated excitons in the lower-n 2D phases are dissociated into free carriers, reducing the useful excitons in the light-emitting phases (the lowest-bandgap phase). Moreover,the loss of excitons should be considered during the inter-phase energy transfer because it does not occur with 100% of quantum efficiency. In the presence of the lower-n 2D phases, the generated excitons should undergo many energy-transfer steps, hampering the maximization of survival of useful excitons at the light-emitting phases. Hence, some substantial approaches to designing an ideal energy landscape having a narrow 2D phase distribution (Figure 1.16c) have been developed, giving a chance to overcome the inherent poor film PLQY of blue color perovskite materials.17Figure 1.15. (a) FRET theory with the distance (d) between the donor and acceptor. Absorption and PL spectra of donor (n = Figure 1.16. (a) Schematic illustration of the formation of exciton-polarons in 2D perovskites. (b) Arrhenius plot of experimental ln(kdis) as a function of negative reciprocal of temperature (–1/T) for n = 2 2D perovskites. (The red dashed line shows the simulated result by using steady-state Eb of 170 meV). Repri18CHAPTER 2. Manipulated Interface for Enhanced Energy Cascade in Quasi-2D Blue Perovskite Light-Emitting DiodesChapter 2 is reproduced in part with permission. Copyright 2022, American Chemical Society.Shin, Y. S.; Park, C. B.; Adhikari, A.; Yoon, Y. J.; Cho, H. W.; Son, J. G.; Seo, J.; Song, T.; Lee, W.;Yeop, J.; Kim, J. W.; Gong, M.; Walker, B.; Kwon, O.-H.; Kim, G.-H.; Kim, J. Y, Manipulated Interface for Enhanced Energy Cascade in Quasi-2D Blue Perovskite Light-Emitting Diodes. ACS Energy Letters 2022, 7 (10), 3345–3352.Contributions:Transient absorption spectroscopy measurements and analysis were conducted by Dr. Aniruddha Adhikari and Prof. Oh-Hoon Kwon.2.1. Research BackgroundRecently, PeLEDs have drawn considerable research interest as attractive optoelectronic devices due to their excellent optoelectrical properties.13,26-28 Significant progress has been made in enhancing the performance of PeLEDs in the past few years,29-32 but the mismatch of energy levels between light-emitting perovskites and adjacent HIL,33,34 as well as poor material stability35-37 have hampered practicaldisplay applications, especially for blue PeLEDs.For efficient blue PeLEDs, quasi-2D perovskites have emerged, taking advantage of their higher exciton-binding energies to confine excitons in the emissive phases.17 Specifically, the thin organic spacer phenethylammonium is typically used to induce and separates 2D phases, consisting of different numbers (n) of sheets of octahedral [PbX6]4– layers.22,38,39 The short distance between the neighboring sheets facilitates sequential Förster-type resonance energy transfer, where energy is transferred from a donor (in this case, low – n sheets of [PbX6]4–) to an acceptor with a lower band gap (in this case, high – n sheets of [PbX6]4–). As a consequence, excitons undergo cascading energy transfer to the lowest-bandgap phases, resulting in efficient light emission of the desired color.40-42 This strategy is advantageous for quasi-2D perovskites, which require an ideal energy landscape to minimize the energy loss of excitons and luminescence of undesired colors.19The next step in maximizing the quantum efficiency of PeLEDs is further minimizing energy losses induced by energetic disorder and defe"}}},{"rowIdx":87,"cells":{"Researcher Name":{"kind":"string","value":"Bowen Jin"},"ORCID":{"kind":"string","value":"-"},"Researcher Query Keyword":{"kind":"string","value":"Enhanced Perovskite Crystals for Radiation Detection"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Thermal Expansion Behavior of Halide Perovskite Single Crystals Across a Broad Temperature Range': 'Title: Thermal Expansion Behavior of Halide Perovskite Single Crystals Across a Broad Temperature Range\\\\nABSTRACT\\\\nHere I present a complete study of lattice dynamics of negative thermal ex-\\\\npansion (NTE) material Scandium trifluoride (ScF3) and discovery of NTE in\\\\nmercurous iodide (Hg2I2). An inelastic x-ray scattering (IXS) study of ScF3\\\\nreveals copious evidence of an incipient soft-mode instability indicating that\\\\nthe T=0 state of ScF3 is very close to a structural quantum phase transition\\\\n(SQPT). Is the anomalously strong and thermally persistent NTE behavior of\\\\nScF3 a consequence of the SQPT? In order to address that, we have explored\\\\nthe coefficient of thermal expansion (CTE) and soft mode behavior of a sec-\\\\nond stoichiometric compound, situated near a SQPT. A detailed side-by-side\\\\ncomparison of the metal trifluorides and mercurous halides suggest strong\\\\nsimilarities and a generic connection between the fluctuating ground state\\\\nof incipient ferroelastic materials and SNTE. We find experimental evidences\\\\nSahan Handunkanda - University of Connecticut - 2018\\\\nfor two- dimensional nanoscale correlations exist at momentum-space regions\\\\nassociated with possibly rigid rotations of the perovskite octahedra of ScF3.\\\\nThe \\\\ndiscussion is extended by addressing the extent to which rigid octahedral\\\\nmotion describes the dynamical fluctuations behind NTE by generalizing a\\\\nsimple model supporting a single floppy mode that is often used to heuris-\\\\ntically describe instances of NTE. Temperature-dependent infrared reflection\\\\nmeasurement on single crystal ScF3 is performed to understand the zone cen-\\\\nter lattice dynamics of ScF3. In addition, I also carried out an instrumentation\\\\ndevelopment project in the laboratory at the Department of Physics which will\\\\nbe briefly discussed in the last chapter.\\\\nLattice Dynamics Studies of Negative\\\\nThermal Expansion Due to Two\\\\nLow-temperature Lattice Instabilities\\\\nSahan Handunkanda\\\\nM.S. Physics, University of Connecticut, Storrs, CT, 2014\\\\nB.S. Physics, University of Colombo, Sri Lanka, 2009\\\\nA Dissertation\\\\nSubmitted in Partial Fulfillment of the\\\\nRequirements for the Degree of\\\\nDoctor of Philosophy\\\\nat the\\\\nUniversity of Connecticut\\\\n2018\\\\nCopyright by\\\\nSahan Handunkanda\\\\n2018\\\\nii\\\\nAPPROVAL PAGE\\\\nDoctor of Philosophy Dissertation\\\\nLattice Dynamics Studies of Negative\\\\nThermal Expansion Due to Two\\\\nLow-temperature Lattice Instabilities\\\\nPresented by\\\\nSahan Handunkanda, M.S. Physics, B.S. Physics\\\\nMajor Advisor\\\\nJason Hancock\\\\nAssociate Advisor\\\\nBarrett Wells\\\\nAssociate Advisor\\\\nDouglas Hamilton\\\\nUniversity of Connecticut\\\\n2018\\\\niii', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation "}}},{"rowIdx":88,"cells":{"Researcher Name":{"kind":"string","value":"Rui Feng"},"ORCID":{"kind":"string","value":"0009-0008-4599-4447"},"Researcher Query Keyword":{"kind":"string","value":"Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":89,"cells":{"Researcher Name":{"kind":"string","value":"Yong-Qi Zhen"},"ORCID":{"kind":"string","value":"0000-0002-1537-7666"},"Researcher Query Keyword":{"kind":"string","value":"Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives': 'Title: Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives\\\\nDesign and synthesis of new1,2,4-oxadiazole/quinazoline-4-one hybrids withantiproliferative activity asmultitargeted inhibitorsAmira M. Mohamed1, Ola M. F. Abou-Ghadir1, Yaser A. Mostafa1,Kholood A. Dahlous2, Stefan Bräse3* and Bahaa G. M. Youssif1*1Pharmaceutical Organic Chemistry Department, Faculty of Pharmacy, Assiut University, Assiut, Egypt,2Department of Chemistry, College of Science, King Saud University, Riyadh, Saudi Arabia, 3Institute ofBiological and Chemical Systems, IBCS-FMS, Karlsruhe Institute of Technology, Karlsruhe, GermanyIntroduction: The combination of BRAF and tyrosine kinase (TK) inhibitors hasbeen demonstrated to be highly effective in inhibiting tumor development and isan approach for overcoming resistance in clinical trials. Accordingly, a novelseries of 1,2,4-oxadiazole/quinazoline-4-one hybrids was developed asantiproliferative multitargeted inhibitors.Methods: The structures of the newly synthesized compounds 9a-o werevalidated using IR, NMR, MS, and elemental techniques. 9a–o were tested asantiproliferative agents.Results and Discussion: The results showed that the majority of the testedcompounds showed significant antiproliferative action with 9b, 9c, 9h, 9k, and9l being the most potent. Compounds 9b, 9c, 9h, 9k, and 9l were tested as EGFRand BRAFV600E inhibitors. These in vitro tests revealed that compounds 9b, 9c, and9h are strong antiproliferative agents that may act as dual EGFR/BRAFV600Einhibitors. 9b, 9c, and 9h were further investigated for their inhibitory effecton mutant EGFR (EGFRT790M), and the results showed that the tested compoundshad considerable inhibitory action. Cell cycle study and apoptosis detectiondemonstrated that compound 9b exhibits cell cycle arrest at the G2/Mtransition. Molecular docking simulations reveal the binding mechanism of themost active antiproliferative agents.KEYWORDSquinazolinone, oxadiazole, kinases, apoptosis, antiproliferative, EGFR, BRAF1 IntroductionDrug developers have spent decades generating selective medicines for specific targets(Medina-Franco et al., 2013; Zhou et al., 2019). Despite the effectiveness of many single-target selective medications, the advancement of multifactorial disorders such as cancer andneurological diseases included many signaling pathways (Fu et al., 2017; Raghavendra et al.,2018). As a result, there is a growing interest in developing treatments that address manytargets at the same time.OPEN ACCESSEDITED BYXuekui Xia,Biology Institute of Shandong Academy ofSciences, ChinaREVIEWED BYChao Liu,Shandong University, ChinaRavikumar Jimmidi,Baylor College of Medicine, United States*CORRESPONDENCEBahaa G. M. Youssif,bgyoussif2@gmail.com,bahaa.youssif@pharm.aun.edu.egStefan Bräse,braese@kit.eduRECEIVED 11 June 2024ACCEPTED 05 August 2024PUBLISHED 30 August 2024CITATIONMohamed AM, Abou-Ghadir OMF, Mostafa YA,Dahlous KA, Bräse S and Youssif BGM (2024)Design and synthesis of new 1,2,4-oxadiazole/quinazoline-4-one hybrids withantiproliferative activity asmultitargeted inhibitors.Front. Chem. 12:1447618.doi: 10.3389/fchem.2024.1447618COPYRIGHT© 2024 Mohamed, Abou-Ghadir, Mostafa,Dahlous, Bräse and Youssif. This is an open-access article distributed under the terms of theCreative Commons Attribution License (CC BY).The use, distribution or reproduction in otherforums is permitted, provided the originalauthor(s) and the copyright owner(s) arecredited and that the original publication in thisjournal is cited, in accordance with acceptedacademic practice. No use, distribution orreproduction is permitted which does notcomply with these terms.Frontiers in Chemistry frontiersin.org01TYPE Original ResearchPUBLISHED 30 August 2024DOI 10.3389/fchem.2024.1447618There are currently two opposing methodologies for designingmulti-targeting medicines. The first technique involves establishingan additive or synergistic effect of various medications operating onseparate targets through combination drug therapy. Preclinicalevidence of enhanced apoptosis and delayed resistance to BRAF(Rapidly Accelerated Fibrosarcoma, B-family) inhibitors (Paraisoet al., 2010; Flaherty et al., 2012), for example, prompted the FDA toapprove a combination of dabrafenib (BRAF inhibitor) andtrametinib (MEK inhibitor) for the treatment of metastaticmelanoma with BRAF mutations (Robert et al., 2015; Wahidet al., 2018). The use of palbociclib and letrozole in the treatmentof advanced breast cancer is another example of successfulcombination therapy (Finn et al., 2016).The second approach is designing and generating multiple-targeting medicines that synergistically block numerouscarcinogenic pathways (Keith et al., 2005; Boran and Iyengar,2010). The method of multi-targeting therapies is finding a singleagent that can operate on two or more targets simultaneously.Cabozantinib, also known as cabometyx, was approved by theFDA as a small molecule dual-targeting inhibitor of the tyrosinekinases c-Met (mesenchymal-epithelial transition factor) andVEGFR-2 (Vascular Endothelial Growth Factor Receptor) andhas been demonstrated to suppress tumor growth, metastasis,and angiogenesis (Food and Administration, 1997).On the other hand, the combination of BRAF and tyrosinekinase (TK) inhibitors has been demonstrated to be highly effectivein inhibiting tumor development and is an approach for overcomingresistance in clinical trials. Vemurafenib (BRAFV600E inhibitor)resistance in thyroid cancer can be addressed by combining itwith EGFR (Epidermal Growth Factor Receptor) inhibitors(Notarangelo et al., 2017). This combination has also shownpromising results in BRAFV600E colorectal cancer (Mondaca et al.,2018). In addition, various compounds have been discovered in vitrothat include the key pharmacophoric groups required to inhibittyrosine kinases, such as EGFR/VEGFR-2 and BRAF (Okaniwaet al., 2012; Zhang et al., 2013). Compound I (Figure 1) inhibitedwild-type BRAF and EGFR with IC50 values in the nanomolar range.Additionally, imidazo [1,2-b]pyridazine II inhibited BRAF andVEGFR-2.Heterocyclic moieties form the largest and most varied class oforganic molecules. In medicinal chemistry, compounds containingheterocyclic nuclei have gained great interest because of their diversetherapeutic actions (Padwa and Bur, 2007). Heterocyclics play acrucial role in the breakdown of all living things and participate invarious biochemical processes necessary for life (Kitadai andMaruyama, 2018). The heteroaromatic framework resemblesbiologically active compounds such as nucleic acids, hormones,and neurotransmitters (Meanwell, 2017). As a result, thesemoieties could be used to design safer medications. Heterocyclesare often found in nature and have been exploited to develop anti-cancer drugs that target many sites and disrupt cancer growthpathways (Sharma et al., 2017). Heterocyclic rings can bemodified with various substituents to cover various chemicals,making them ideal for designing anti-cancer drugs.Nitrogen-containing heterocyclic chemicals significantly affectabout 75% of FDA-approved anti-cancer drugs (Kerru et al., 2020).Quinazolinone, a bicyclic system composed of benzene andpyrimidinone rings, is one of the most common nitrogen-containing heterocycles in medicinal chemistry, found in variouscompounds with diverse biological activity. Idelalisib III (Do et al.,2016), Ispinesib IV (Purcell et al., 2010), and Halofuginone V(Figure 2) (Derbyshire et al., 2012; McLaughlin et al., 2014) areexamples of recently approved or marketed medications with anti-cancer properties.Depending on the position of the keto or oxo group, threedistinct forms are possible: quinazolin-2(1H)-one VI, quinazoline-2,4-(1H,3H)-di-one VII, and quinazolin-4(3H)-one VIII (Figure 3).Among these, quinazolin-4-one VIII is the most commonly usedscaffold in synthetic processes or as a structural component ofnatural compounds (Sharma et al., 2017). This last scaffold isadaptable, allowing up to six potential substitutes in positions 2,3, 5, 6, 7, and 8.In recent publications (Hisham et al., 2022; Hisham et al., 2023),we present the design and synthesis of a new series of quinazoline-4-one/chalcone hybrids that function as dual inhibitors of EGFR andBRAFV600E with antiproliferative activity. The target compoundswere tested in vitro against various cancer cell lines and the EGFRand BRAFV600E enzymes. Compound IX (Figure 4) was the mostpotent derivative, with a GI50 of 1.16 µM, compared to the referencedrug Doxorubicin (GI50 = 1.14 µM). Compound IX showedsignificant inhibitory activity against EGFR and BRAFV600E, withIC50 values of 0.11 µM and 0.65 µM, respectively. Moreover,apoptosis assay results revealed that compound IX enhanced thelevel of active caspase-3, 8, and 9 with significant induction ofcytochrome c and Bax levels and downregulation of the anti-apoptotic Bcl-2 levels.FIGURE 1Structures of compounds I and II.Frontiers in Chemistry frontiersin.org02Mohamed et al. 10.3389/fchem.2024.1447618On the other hand, literature reviews reveal that 1,2,4-oxadiazoles have statistical significance in bioorganic andmedicinal chemistry. They are recognized for their diversepharmacological characteristics (Benassi et al., 2020; El Mansouriet al., 2020; Loboda et al., 2020). The 1,2,4-oxadiazole exhibitsbioisosteric equivalence with ester and amide moieties. Whenunstable conditions (e.g., hydrolysis) are identified, 1,2,4-oxadiazole is a highly effective alternative (Hendawy, 2022). Thesubstantial biological impact of 1,2,4-oxadiazole derivatives oncancer cells can be attributed to various mechanisms of action.For example, we developed and synthesized novel 1,2,4-oxadiazole-based derivatives linked to a triaryl-imidazole moiety, withcompound X (Figure 4) being the most potent (Youssif et al.,2022). In vitro studies assessed the antiproliferative effects ofrecently identified compounds inhibiting p38α and BRAFV600E.These compounds showed effective antiproliferative and kinaseinhibition.Another set of 1,2,4-oxadiazole-based compounds (compoundXI, Figure 4) were designed, synthesized, and tested forantiproliferative properties against EGFR-TK. The experimentshowed promising antiproliferative effects against cancer celllines, with low micromolar IC50 values against EGFR, comparedto the reference doxorubicin (Unadkat et al., 2021).1.1 Rational designConsistent with prior findings and continuing our efforts todevelop dual or multitargeted antiproliferative agents (Al-Wahaibiet al., 2020; Alshammari et al., 2022; Al-Wahaibi et al., 2023a; Abdel-Aziz et al., 2023; Al-Wahaibi et al., 2023b; Al-Wahaibi et al., 2023c;Al-Wahaibi et al., 2023d; Al-Wahaibi et al., 2023e), this study’sstrategy was to design and synthesize new antiproliferative agentsbased on quinazoline-4-one/1,2,4-oxadiazole hybrids (Figure 5) toobtain new anti-tumor agents with synergistic activity.The substitutions on the nitrogen atom of the quinazolinemoiety were changed between alkyl (methyl, ethyl, and allyl) andaryl (phenyl or tolyl) moieties to examine the impact of rigidity onthe antiproliferative activity of these compounds. In addition,different substituents, such as a chlorine atom (electronFIGURE 2Examples of approved or commercialized anti-cancer medicines with the quinazoline-4-one scaffold.FIGURE 3Different forms of quinazolinones.Frontiers in Chemistry frontiersin.org03Mohamed et al. 10.3389/fchem.2024.1447618withdrawing group) or a methoxy group (electron donating group),were used to investigate the different electronic impacts of thesesubstituents on 9a-o’s antiproliferative activity.All of the novel analogs were examined for cell viability effectagainst a normal human mammary gland epithelial (MCF-10A) cellline as well as for antiproliferative activity against four human cancercell lines: colon (HT-29), pancreatic (Panc-1), lung (A-549), andbreast (MCF-7). Furthermore, the compounds with the highestantiproliferative activity were investigated in vitro as multi-targeting inhibitors of EGFR, EGFRT790M, and BRAFV600Eenzymes. The study was expanded to include one of the mostactive derivatives, 9b, as a representative agent to evaluate itsFIGURE 4Structures of quinazoline-4-one and 1,2,4-oxadiazole-based derivatives as anticancer agents.FIGURE 5Structures of new targets 9a-o.Frontiers in Chemistry frontiersin.org04Mohamed et al. 10.3389/fchem.2024.1447618mechanistic effects on the cell cycle and induction of apoptosis.Finally, docking studies were performed on the most activecompounds against the selected enzymes to explain their in vitroresults. Furthermore, the ADME analyses were performed toinvestigate their pharmacokinetic features.2 Results and discussion2.1 ChemistryScheme 1 summarizes the synthetic pathways of the new targetcompounds 9a-o. Anthranilic acid (1) was refluxed in ethanol withisothiocyanate derivatives 2a-e for 8 h. After the reaction wascompleted (as determined by TLC), the resulting whiteprecipitate was collected by filtration and recrystallized from anethanol/dioxane mixture (1:1) to give the corresponding quinazolinederivatives 3a-e in 90%–95% yields (Moussa et al., 2018). On theother hand, compounds 6a-c, amidoxime derivatives, weresynthesized in 50%–60% yields over two steps, Scheme 2. Thefirst step involved reacting the corresponding aldehydes 4a-c with28% aqueous ammonia and iodine in THF for 2–3 h to yield theSCHEME 1Synthesis of the new target compounds 9a-oSCHEME 2Structure Activity Relationship (SAR) analysis of new targets 9a-o.Frontiers in Chemistry frontiersin.org05Mohamed et al. 10.3389/fchem.2024.1447618corresponding aryl nitrile derivatives 5a-c in 76%–80% (Yan et al.,2017). The second step was a 12- to 18-h methanol reflux ofcompounds 5a-c with hydroxylamine chloride and sodiumcarbonate (Youssif et al., 2019). Compounds 6a-c were reactedwith chloroacetyl chloride in dry acetone to yieldbenzimidamides (7a-c), which were cyclized by refluxing intoluene to the corresponding 3-aryl-5-(chloromethyl)-1,2,4-oxadiazole derivatives 8a-c as a yellow oil. Compounds 8a-c werepurified using column chromatography with hexane: ethyl acetate(9:1) as an eluent (Minin et al., 2023). For example, the 1H NMRspectrum of compound 8b confirmed the disappearance of twoprotons from the NH2 group of the corresponding amidoxime 7b.Moreover, the spectrum displayed a singlet signal corresponding tothe methylene protons (Cl-CH2) at δ 4.74. The spectra also revealeda characteristic pair of doublets in the aromatic region for 4-ClC6H4at δ 8.01 and 7.46.Reagents and Conditions: a) Triethylamine, ethanol, Reflux 8 h;b) ammonia (28%), I2, THF, Stirring 1h; c) NH2OHHCl, Na2CO3,THF, Reflux 18 h; d) Chloroacetyl Chloride, K2CO3, Dry acetone,Stirring 24hrs; e) Toluene, Reflux 10 h; f) K2CO3, KI, DMF,Stirring 24 h.Finally, the target novel compounds, 9a-o, were synthesizedin high yields by coupling compounds 3a-e with thecorresponding 1,2,4-oxadiazoles 8a-c in DMF using anhydrousK2CO3 and KI and stirring for 24 h at room temperature. 9a-owere purified via ethanol recrystallization. The structures of 9a-owere elucidated using 1H NMR, 13C NMR, and elementalmicroanalyses. The 1H NMR spectra of compound 9l, as anexample, confirmed the presence of ethyl group characteristicsignals in the form of triplet at δ 1.31 (t, J = 7.1 Hz, 3H,N-CH2CH3) and quartet at δ 4.11 (q, J = 7.1 Hz, 2H, N-CH2).The spectrum also revealed two distinct singlet signals: at δ 3.81(s, 3H, OCH3) and at δ 4.91 (s, 2H, S-CH2). Additionally, thespectrum revealed a pair of doublets of the aromatic ring’s para disubstitution pattern and extra signals for the aromatic protons inthe quinazoline moiety. The 13C NMR spectrum of 9l indicatedthe presence of ethyl group characteristic signals at δ 39.56 and δ13.01, methylene group at δ 26.62, and methoxy group at δ 55.TABLE 1 Cell viability percent and antiproliferative activity (IC50 values) of compounds 9a-o.Comp Cell viability % R1 R2 Antiproliferative activity IC50 ± SEM (nM)A-549 MCF-7 Panc-1 HT-29 Average (GI50)9a 90 Phenyl H 46 ± 4 50 ± 4 48 ± 4 48 ± 4 489b 89 Phenyl Cl 22 ± 2 26 ± 2 24 ± 2 24 ± 2 249c 91 Phenyl OMe 24 ± 2 28 ± 3 26 ± 2 25 ± 2 269d 90 p-Tolyl H 54 ± 4 58 ± 5 54 ± 5 55 ± 5 559e 91 p-Tolyl Cl 40 ± 3 44 ± 4 42 ± 4 42 ± 4 429f 92 p-Tolyl OMe 50 ± 4 53 ± 5 52 ± 5 52 ± 5 529g 90 m-Tolyl H 49 ± 4 52 ± 5 50 ± 5 50 ± 5 509h 89 m-Tolyl Cl 28 ± 2 31 ± 3 29 ± 2 30 ± 3 309i 91 m-Tolyl OMe 65 ± 6 69 ± 6 66 ± 6 68 ± 6 679j 90 Ethyl H 69 ± 6 76 ± 7 68 ± 6 68 ± 6 709k 93 Ethyl Cl 32 ± 3 35 ± 3 34 ± 3 34 ± 3 349l 90 Ethyl OMe 36 ± 3 40 ± 4 38 ± 3 37 ± 3 389m 92 Allyl H 62 ± 6 65 ± 6 64 ± 6 62 ± 6 639n 91 Allyl Cl 43 ± 4 47 ± 4 45 ± 4 44 ± 4 459o 89 Allyl OMe 56 ± 4 59 ± 5 57 ± 5 57 ± 5 57Erlotinib ND — — 30 ± 3 40 ± 3 30 ± 3 30 ± 3 33ND: not determined.Frontiers in Chemistry frontiersin.org06Mohamed et al. 10.3389/fchem.2024.1447618Elemental microanalysis of 9l confirmed that the calculated data(%) were C, 60.90; H, 4.60; N, 14.20; S, 8.13, while the found data(%) were C, 61.13; H, 4.74; N, 14.37; S, 8.20.2.2 Biology2.2.1 Assay of cell viability effectThe human mammary gland epithelial (MCF-10A) normal cellline was used to test the viability of novel compounds 9a-o(Mahmoud et al., 2022; Mekheimer et al., 2022). After 4 days ofincubation on MCF-10A cells, the cell viability of compounds 9a-owas determined using the MTT test. Table 1 demonstrates that noneof the compounds examined were cytotoxic, and all hybrids showedmore than 89% cell viability at 50 µM.2.2.2 Assay of antiproliferative actionThe MTT assay was used to investigate the antiproliferativeactivity of hybrids 9a-o versus four human cancer cell lines usingErlotinib as a control: colon cancer (HT-29) cell line, pancreaticcancer (Panc-1) cell line, lung cancer (A-549) cell line, and breastcancer (MCF-7) cell line (El-Sherief et al., 2019; Al-Wahaibi et al.,2022). Table 1 displays the median inhibitory concentration (IC50)and GI50 (average IC50) against the four cancer cell lines.In general, the hybrids 9a-o had significant antiproliferative actionwith GI50 values ranging from 24 nM to 70 nM versus the four cancercell lines evaluated, compared to Erlotinib, which had a GI50 value of33 nM. Compounds 9b, 9c, 9h, 9k, and 9l were the most potent fivederivatives, with GI50 values of 24, 26, 30, 34, and 38 nM,making 9b, 9c,and 9hmore potent than Erlotinib (GI50 = 33 nM). Out of all the newlysynthesized hybrids 9a-o, compound 9b (R1 = phenyl, R2 = Cl) had thehighest potency, with a GI50 value of 24 nM, which was 1.4 times morepotent than the reference Erlotinib (GI50 = 33 nM).The type of the aryl/alkyl moieties at position 3 of thequinazoline moiety appears to be critical for 9a-o hybridsantiproliferative activity. The GI50 value of compound 9h (R1 =m-tolyl, R2 = Cl) was 30 nM, less potent than compound 9b but stillmore potent than the reference erlotinib (GI50 = 33 nM).Moreover, Compounds 9e (R1 = p-tolyl, R2 = Cl), 9k (R1 = ethyl,R2 = Cl), and 9n (R1 = allyl, R2 = Cl) demonstrated GI50 values of 42,34, and 45 nM, respectively, being less potent than compounds 9b,9h, and even Erlotinib. These results indicated the importance of thequinazoline moiety position three substitution pattern onantiproliferative activity, with activity rising in the followingorder: phenyl > m-tolyl > p-tolyl > ethyl > allyl.Compound 9c (R1 = phenyl, R2 = OMe) rated second in activityagainst the four cancer cell lines, with a GI50 value of 26 nM, slightlyless effective than 9b but still more potent than Erlotinib (GI50 =33 nM). The unsubstituted phenyl derivative, 9a (R1 = phenyl, R2 =H), was less potent than 9b and 9c, indicating that the substitutionpattern at the fourth position of the phenyl group in the 1,2,4-oxadiazole moiety affects the antiproliferative activity of thesehybrids, with activity increasing in the order Cl > OMe > H.Regardless of the nature of the substitution pattern at position3 of the quinazoline moiety, the same rule (Cl > OMe > H inactivity) applies to other derivatives.2.2.3 EGFR inhibitory assayThe EGFR-TK test (Abdel-Aziz et al., 2023) was used to assessthe inhibitory potency of the most effective antiproliferativederivatives 9b, 9c, 9h, 9k, and 9l against EGFR, and the resultsare shown in Table 2. This assay’s results are in line with theantiproliferative assay’s, which found that compounds 9b (R1 =phenyl, R2 = Cl), 9c (R1 = phenyl, R2 = OMe), and 9h (R = m-tolyl,R2 = Cl), the most effective antiproliferative hybrids, were the mostpotent derivatives of EGFR inhibitors, with IC50 values of 57 ± 4 nM,64 ± 5 nM, and 72 ± 5 nM, respectively, surpassing the referencedrug Erlotinib (IC50 = 80 ± 5). Compounds 9k (R1 = ethyl, R2 = Cl)and 9l (R1 = ethyl, R2 = OMe) demonstrated significant anti-EGFRactivity, with IC50 values of 84 ± 6 nM and 91 ± 07 nM, respectively,which were less potent than Erlotinib. These findings show thatcompounds 9b, 9c, and 9h have significant EGFR inhibitory actionand could be used as antiproliferative agents.2.2.4 BRAFV600E inhibitory assayAn in vitro investigation assessed the anti-BRAFV600E activity of9b, 9c, 9h, 9k, and 9l (Youssif et al., 2022). The enzyme assaydemonstrated that the five hybrids examined substantially inhibitedBRAFV600E, with IC50 values ranging from 48 to 70 nM, Table 2. In allcases, the IC50 of the examined compounds is greater than that of thereference Vemurafenib (IC50 = 30). Compounds 9b, 9c, and 9hdemonstrated the most effective inhibitory activity againstTABLE 2 IC50 values of compounds 9b, 9c, 9h, 9k, and 9l against EGFR, BRAFV600E and EGFRT790M.Compound EGFR inhibitionIC50 ± SEM (nM)BRAFV600E inhibition IC50 ± SEM (nM) EGFRT790M inhibition IC50 ± SEM (nM)9b 57 ± 4 48 ± 4 10 ± 19c 64 ± 5 51 ± 5 11 ± 29h 72 ± 5 57 ± 5 15 ± 29k 84 ± 6 62 ± 5 ND9l 91 ± 7 70 ± 5 NDErlotinib 80 ± 5 60 ± 5 NDVemurafenib ND 30 ± 3 NDOsimertinib ND ND 8 ± 1ND: not determined.Frontiers in Chemistry frontiersin.org07Mohamed et al. 10.3389/fchem.2024.1447618BRAFV600E (IC50 = 48, 51, and 57 nM, respectively) and werediscovered to be potent inhibitors of cancer cell growth (GI50 =24, 26, and 30 nM, respectively). As a result, compounds 9b, 9c, and9h are effective antiproliferative agents that function as dual EGFR/BRAFV600E inhibitors.2.2.5 EGFRT790M inhibitory assayThe HTRF KinEASE-TK assay (Miles et al., 2020) was used toevaluate the inhibitory action of the most potent hybrids, 9b, 9c, and9h, against mutant-type EGFR (EGFRT790M). As demonstrated inTable 2, compounds 9b, 9c, and 9h displayed excellent inhibitoryeffect against EGFRT790M, with IC50 values of 10 ± 1, 11 ± 1, and 15 ±1 nM, respectively, being equivalent to the reference Osimertinib(IC50 = 8 ± 1 nM), which may explain their robust antiproliferativeactivity. These findings suggested that phenyl and m-tolylsubstitutions in the quinazoline moiety’s third position, as well aschlorine atom and methoxy substitutions in the para-position of thephenyl group in the 1,2,4-oxadiazole moiety, are required for theinhibitory impact on EGFR, BRAFV600E, and EGFRT790M.1. The type of the aryl/alkyl moieties at position 3 of thequinazoline moiety appears to be critical for 9a-o hybridsantiproliferative activity, with activity increasing in thefollowing order: phenyl > m-tolyl > p-tolyl > ethyl > allyl.2. The substitution pattern at the fourth position of the phenylgroup in the 1,2,4-oxadiazole moiety affects theantiproliferative activity of these hybrids as well as EGFR,BRAFV600E, and EGFRT790M inhibition, with activity increasingin the order Cl > OMe > H.3. Regardless of the nature of the substitution pattern at position3 of the quinazoline moiety, the same rule (Cl > OMe > H inactivity) applies to other derivatives.2.2.6 Cell cycle analysis and apoptosis assays2.2.6.1 Cell cycle analysisCompound 9b was investigated for its effects on cell cycleprogression and apoptosis induction in A-549 cells. A lungcancer (A-549) cell line was treated for 24 h with an IC50concentration of 9b. The cell line was labeled with PI/Annexin V,and flow cytometry was done with a BD FASC Caliber (El-Sheriefet al., 2018). The results (Figure 6) showed that A-549 treated withcompound 9b had a significant percentage of cell accumulation(29%) in the G2/M phase after 24 h of incubation, indicating cellcycle arrest at the G2/M transition.2.2.6.2 Apoptosis induction assayTo assess 9b′s potential to induce apoptosis, A-549 cells werelabeled with Annexin V/PI, grown for 24 h, and evaluated.Examining early and late apoptosis demonstrated that compound9b could produce high levels of apoptosis, with a necrosis percentage6.43 (Figures 7, 8).2.3 Docking simulationsStarting with RSCB deposited crystal structure of EGFR proteinhaving Erlotinib as a co-crystallized ligand (PDB: 1M17) (Bhat et al.,2022), and re-docking of Erlotinib revealed a docking score (S)of −7.30 kcal/mol and an RMSD of 1.28 Å, in addition to formationof the two characteristic H-bond interactions with two of key aminoacid residues, Gln767 and Met769, indicating validity of dockingstudy parameters. While running docking simulations within EGFRactive site (PDB ID: 1M17) showed that most of the test derivatives(9a-o) showed moderate to strong binding interactions(S = −5.93 to −7.52; c.f. Erlotinib: 7.30 kcal/mol) as listed inTable 3. These interactions were variable between H-bond and/orH-pi, with key amino acid residues (Met 769, Lys 721, Gly 772, andLeu 694) lining the active site, as shown in Supplementary Figure S1(Supplementary Material).Remarkably, derivative 9b (R1 = Ph. and R2 = Cl) showed ahigher docking score (S = −6.51 kcal/mol) over its methoxycongener (S = −6.04 kcal/mol) and has the best docking scoreover other p-chloro derivatives (9e, 9h, and 9k). Visualinspection of the docking pose with the lowest RMSD value andhighest docking score of compound 9b, we observed a stabilizingH-bonding and H-pi binding interactions through N-phenyl andamidic carbonyl group of quinazoline ring with Lys721 amino acidresidue, Figure 9. Such interactions were not found in other p-chloroFIGURE 6Cell cycle analysis results for compound 9b.FIGURE 7Apoptosis induction results of compound 9b.Frontiers in Chemistry frontiersin.org08Mohamed et al. 10.3389/fchem.2024.1447618derivatives, 9e and 9k (except with derivative 9h), because of thehydrophilic tale of Gln738 amino acid residue, which repels closelyfound non-hydrophilic groups as methyl group of 9e and ethylgroup of 9k, as shown in Supplementary Figure S1.Whereas upon working on binding interactions within mutantEGFR (EGFRT790M; PDB ID: 2JIU) active site, the p-Cl derivative(9b) gave the highest docking score (S = −7.43 kcal/mol) among all15 derivatives tested, as shown in Table 3. Most test derivativescommonly interacted through H-bonds and/or H-pi interactionswith Lys 745 and Leu 718, as seen with docking poses of derivatives9b and 9h, Figure 10.Finally, docking scores of derivatives (9a-o) interactions withinBRAF (PDB ID: 5JRQ) (Umar et al., 2020) active site were high andso close to each other (S = −6.24 to −7.65 kcal/mol). Additionally,multiple interactions varying from the weak Pi-Pi interactions,through H-Pi to the strong H-bonds with either Phe 583, Val471, Asp 594, or Lys 483, as shown in Figure 11 of compound 9l.To summarize, all the 15 derivatives showed good bindingprofiles within target proteins EGFR, mutant EGFR (EGFRT790M),and BRAFV600E, as seen from their docking scores and interactionswith amino acid residues lining their active sites, and this could beused to explain the possible mechanism by which such class ofcompounds inhibit these proteins activity.2.4 Calculation of ADME propertiesThe drug-likeness of new compounds 9a-o was calculated usingthe SwissADME website (Daina et al., 2017) to predict theirtransport properties through membranes like GIT and/or BBB.All the test compounds obey Lipinski’s rule of five (RO5) withMLogP below 5, in addition to having both a topological polarsurface area below 140 Å2 and molar refractivity below 130,indicating their facile transport through cell membranes andhence better oral bioavailability (F), as shown in Table 4.3 ConclusionA novel set of quinazoline-4-one/1,2,4-oxadiazole hybrids (9a-o)was designed and synthesized as EGFR, EGFRT790M, and BRAFV600Einhibitors in the search for multitargeted antiproliferative scaffold.The novel hybrids showed encouraging antiproliferative actions.Compounds 9b, 9c, 9h, 9k, and 9l were evaluated as EGFR andBRAFV600E inhibitors. These in vitro experiments demonstrated thatcompounds 9b, 9c, and 9h are potent antiproliferative agents capableof acting as dual EGFR/BRAFV600E inhibitors. 9b, 9c, and 9h werefurther studied for their inhibitory effect on mutant EGFR(EGFRT790M), with the results indicating that the evaluatedcompounds had a significant inhibitory effect. Cell cycle analysisand apoptosis induction assay of 9b revealed cell cycle arrest at the G2/M phase, which can induce apoptosis. EGFR and BRAFV600E dockingsimulations inside their active regions shed light on these compounds’possible modes of inhibition. ADME calculations revealed that all testcompounds satisfy Lipinski’s rule of five (RO5) with MLogP <5, witheasy transport through cell membranes and higher oral bioavailability.These new hybrids may have potential as anti-cancer drugs afteroptimization.FIGURE 8Cell cycle analysis and apoptosis induction results of compound 9b.Frontiers in Chemistry frontiersin.org09Mohamed et al. 10.3389/fchem.2024.14476184 Experimental4.1 Chemistry4.1.1 General detailsThe starting materials, quinazolines 3a-e (Moussa et al., 2018)and 3-aryl-5-(chloromethyl)-1,2,4-oxadiazole derivatives,compounds 8a-c (Minin et al., 2023), were prepared according toliterature methods.4.1.2 General procedures for the synthesis ofcompounds (9a-o)To a stirred solution of quinazoline derivatives (0.60 mmol,1 eq), compounds 3a-e, in DMF (5 mL), anhydrous K2CO3(0.72 mmol, 1.20 eq, 0.10 g) was added and stirred for 1h atroom temperature. Then, 3-aryl-5-(chloromethyl)-1,2,4-oxadiazole derivatives, compounds 8a-c, (0.60 mmol, 1 eq.)was added, and KI (0.60 mmol, 1 eq, 0.10 g) was also addedto the reaction mixture and stirring was continued for 24 h. Aftercompletion of the reaction (checked by TLC using Hexane: Ethylacetate 2:1), the reaction mixture was poured into crushed icewhile stirring. The obtained precipitate was filtered off, washedseveral times with water, dried at 60°C, and crystallizedfrom ethanol.4.1.3 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-phenylquinazolin-4(3H)-one (9a)Yield: 0.21 g (85%), White solid, m.p: 162°C–164°C, Rf. 0.66(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.07 (d, J = 7.7 Hz, 1H, Ar-H), 7.99 (d, J = 6.4 Hz,2H, Ar-H), 7.81 (t, J = 7.4 Hz, 1H, Ar-H), 7.60 (d, J = 11.0 Hz,4H, Ar-H), 7.57–7.52 (m, 4H, Ar-H), 7.47 (t, J = 8.0 Hz, 2H, Ar-H), 4.79 (s, 2H, S-CH2);13C NMR (100 MHz, δ ppm DMSO-d6):177.09, 167.84, 160.56, 155.46, 146.85, 135.49, 135.01, 131.59,130.22, 129.67, 129.46, 129.25, 126.98, 126.61, 126.32, 126.03,125.90, 119.60, 26.97; Anal. Calc. (%) For C23H16N4O2S: C,66.97; H, 3.91; N, 13.58; S, 7.77. Found: C, 66.81; H, 3.85; N,13.82; S, 7.85.4.1.4 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3 phenylquinazolin-4(3H)-one (9b)Yield: 0.24 g (89%), White solid, m.p: 172°C–174°C, Rf. 0.67(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.07 (dd, J = 8.2, 1.5 Hz, 1H, Ar-H), 8.00 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.84–7.79 (m, 1H, Ar-H), 7.63 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.62–7.60 (m, 3H, Ar-H), 7.57–7.52 (m, 2H, Ar-H),7.51–7.44 (m, 2H, Ar-H), 4.78 (s, 2H, S-CH2);13C NMR (100MHz, δppm DMSO-d6): 177.84, 167.52, 161.02, 155.92, 147.31, 136.82,135.95, 135.47, 130.70, 130.14, 129.93, 129.89, 129.24, 127.08,TABLE 3 Oxadiazoles docking scores in EGFR, EGFRT790M, and BRAFV600E active sites.Comp R1 R2 Docking score (S; kcal/mol)EGFR (1M17) EGFRT790M (2JIU) BRAFV600E (5JRQ)9a Phenyl H −6.0118 −7.1900 −7.04589b Phenyl Cl −6.5073 −7.4268 −7.09859c Phenyl OMe −6.0388 −7.2876 −7.22789d p-Tolyl H −6.0314 −6.7971 −6.92329e p-Tolyl Cl −6.4310 −5.6713 −7.30789f p-Tolyl OMe −6.8251 −6.7228 −6.99719g m-Tolyl H −7.5281 −6.8538 −6.86469h m-Tolyl Cl −6.2685 −6.4894 −6.84709i m-Tolyl OMe −6.8622 −6.7373 −7.65169j Ethyl H −6.6348 −6.9894 −6.28299k Ethyl Cl −5.7431 −6.3041 −6.58249l Ethyl OMe −5.9343 −6.1808 −6.70299m Allyl H −6.1516 −7.2311 −6.83909n Allyl Cl −6.2599 −6.3498 −6.84759o Allyl OMe −6.7193 −6.7397 −6.8732Erlotinib −7.3034 — —Osimertinib — −8.5638 —Vemurafenib — --- −9.3319Frontiers in Chemistry frontiersin.org10Mohamed et al. 10.3389/fchem.2024.1447618126.79, 126.36, 125.33, 120.08, 27.44; Anal. Calc. (%) ForC23H15ClN4O2S: C, 61.81; H, 3.38; N, 12.54; S, 7.17. Found: C,61.97; H, 3.50; N, 12.71; S, 7.28.4.1.5 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-phenylquinazolin-4(3H)-one (9c)Yield: 0.23 g (88%), White solid, m.p: 186°C–188°C, Rf.0.65 Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.07 (d, J = 7.7 Hz, 1H, Ar-H), 7.92 (d, J = 8.7 Hz, 2H,Ar-H p-OCH3 C6H4), 7.81 (t, J = 7.3 Hz, 1H, Ar-H), 7.62 (d, J =3.9 Hz, 3H, Ar-H), 7.59–7.35 (m, 4H, Ar-H), 7.09 (d, J = 8.7 Hz, 2H,Ar-H p-OCH3 C6H4), 4.76 (s, 2H, S-CH2), 3.82 (s, 3H, OCH3);13CNMR (100 MHz, δ ppm DMSO-d6): 176.64, 167.52, 161.71, 160.54,155.46, 146.85, 135.48, 134.99, 130.20, 129.64, 129.45, 128.67,126.60, 126.30, 125.89, 119.60, 118.28, 114.63, 55.36, 26.89; Anal.Calc. (%) For C24H18N4O3S: C, 65.14; H, 4.10; N, 12.66; S, 7.25.Found: C, 64.91; H, 4.27; N, 12.89; S, 7.23.4.1.6 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-p-tolylquinazolin-4(3H)-one (9d)Yield: 0.22 g (86%), White solid, m.p: 168°C–170°C, Rf. 0.69(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.06 (d, J = 7.5 Hz, 1H, Ar-H), 7.98 (d, J = 6.5 Hz, 2H, Ar-H),7.81 (t, J = 7.4 Hz, 3H, Ar-H), 7.59–7.52 (m, 4H, Ar-H), 7.47 (d, J =6.1 Hz, 1H, Ar-H), 7.39 (q, J = 8.2 Hz, 2H, Ar-H), 4.76 (s, 2H,S-CH2), 2.42 (s, 3H, CH3);13C NMR (100 MHz, δ ppm DMSO-d6):177.12, 167.83, 160.62, 155.78, 146.87, 139.99, 135.01, 132.86,131.63, 130.16, 129.29, 129.18, 126.99, 126.63, 126.32, 126.03,125.89, 119.60, 26.96, 20.89; Anal. Calc. (%) For C24H18N4O2S:C, 67.59; H, 4.25; N, 13.14; S, 7.52. Found: C, 67.34; H, 4.43; N,13.40; S, 7.68.4.1.7 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-p-tolylquinazolin-4(3H)-one (9e)Yield: 0.25 g (90%), White solid, m.p: 156°C–158°C, Rf. 0.70(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.05 (d, J = 6.9 Hz, 1H, Ar-H), 7.98 (d, J = 7.5 Hz, 2H, Ar-H p-ClC6H4), 7.80 (t, J = 6.4 Hz, 1H, Ar-H), 7.62 (d, J = 7.5 Hz, 2H, Ar-Hp-Cl C6H4), 7.45 (d, J = 7.5 Hz, 2H, Ar-H), 7.42–7.31 (m, 4H, Ar-H),4.76 (s, 2H, S-CH2), 2.42 (s, 3H, CH3);13C NMR (100 MHz, δ ppmDMSO-d6): 177.42, 167.06, 160.62, 155.76, 146.86, 140.00, 136.36,135.02, 132.84, 130.17, 129.47, 129.17, 128.80, 126.62, 126.34,125.89, 124.87, 119.59, 26.99, 20.90; Anal. Calc. (%) ForC24H17ClN4O2S: C, 62.54; H, 3.72; N, 12.16; S, 6.96. Found: C,62.37; H, 3.80; N, 12.42; S, 6.89.4.1.8 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-p-tolylquinazolin-4(3H)-one (9f)Yield: 0.24 g (89%), White solid, m.p: 165°C–167°C, Rf. 0.68(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.06 (d, J = 7.6 Hz, 1H, Ar-H), 7.91 (d, J = 8.5 Hz, 2H,Ar-H p-OCH3 C6H4), 7.80 (t, J = 7.4 Hz, 1H, Ar-H), 7.48 (d, J =7.8 Hz, 2H, Ar-H), 7.39 (q, J = 7.7 Hz, 4H, Ar-H), 7.09 (d, J =8.5 Hz, 2H, Ar-H p-OCH3 C6H4), 4.74 (s, 2H, S-CH2), 3.82 (s, 3H,OCH3), 2.42 (s, 3H, CH3);13C NMR (100 MHz, δ ppm DMSO-d6): 176.65, 167.54, 161.73, 160.62, 155.74, 146.87, 139.98, 134.97,FIGURE 9Binding Interactions of 9b within EGFR (PDB ID: 1M17) active site showing both H-bonds and H-Pi interactions as green-dotted arrows and lines,respectively.Frontiers in Chemistry frontiersin.org11Mohamed et al. 10.3389/fchem.2024.1447618132.86, 130.15, 129.17, 128.68, 126.62, 126.29, 125.90, 119.59,118.30, 114.65, 55.38, 26.90, 20.90; Anal. Calc. (%) ForC25H20N4O3S: C, 65.77; H, 4.42; N, 12.27; S, 7.02. Found: C,65.62; H, 4.61; N, 12.41; S, 6.98.4.1.9 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-m-tolylquinazolin-4(3H)-one (9g)Yield: 0.22 g (86%), White solid, m.p: 170°C–172°C, Rf. 0.69(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-FIGURE 10Docking poses of compound 9b (top figure) and 9h (bottom figure) within the active site of EGFRT790M (PDB ID: 2JIU) showing H-Pi interactions withkey amino acid residues.Frontiers in Chemistry frontiersin.org12Mohamed et al. 10.3389/fchem.2024.1447618d6): 8.06 (dd, J = 8.2, 1.4 Hz, 1H, Ar-H), 7.98 (dd, J = 7.9, 1.7 Hz, 2H,Ar-H), 7.82–7.78 (m, 1H, Ar-H), 7.59–7.54 (m, 3H, Ar-H),7.49–7.45 (m, 3H, Ar-H), 7.41 (d, J = 7.7 Hz, 1H, Ar-H), 7.31 (d,J = 8.7 Hz, 2H, Ar-H), 4.76 (s, 2H, S-CH2), 2.40 (s, 3H, CH3);13CNMR (100 MHz, δ ppm DMSO-d6): 177.15, 167.86, 160.57, 155.56,146.87, 139.34, 135.40, 135.07, 131.65, 130.91, 129.65, 129.49,129.31, 127.01, 126.63, 126.44, 126.38, 126.04, 125.92, 119.60,26.99, 20.79; Anal. Calc. (%) For C24H18N4O2S: C, 67.59; H, 4.25;N, 13.14; S, 7.52. Found: C, 67.36; H, 4.09; N, 13.41; S, 7.60.4.1.10 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-m-tolylquinazolin-4(3H)-one (9h)Yield: 0.25 g (90%), White solid, m.p: 178°C–180°C, Rf. 0.70(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.09 (d, J = 7.7 Hz, 1H, Ar-H), 8.02 (d, J = 8.6 Hz, 2H,Ar-H p-Cl C6H4), 7.84 (t, J = 8.4 Hz, 1H, Ar-H), 7.66 (d, J =8.6 Hz, 2H, Ar-H p-Cl C6H4), 7.60–7.41 (m, 4H, Ar-H), 7.34 (d,J = 10.4 Hz, 2H, Ar-H), 4.79 (s, 2H, S-CH2), 2.43 (s, 3H, CH3);13CNMR (100 MHz, δ ppm DMSO-d6): 177.37, 167.03, 160.50,155.49, 146.81, 139.28, 136.33, 135.35, 134.99, 130.86, 129.61,129.43, 129.23, 128.76, 126.58, 126.40, 126.31, 125.86, 124.85,119.57, 26.95, 20.75; Anal. Calc. (%) For C24H17ClN4O2S: C, 62.54; H, 3.72; N, 12.16; S, 6.96. Found: C, 62.39; H, 3.85; N,12.40; S, 6.89.4.1.11 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-m-tolylquinazolin-4(3H)-one (9i)Yield: 0.24 g (89%), White solid, m.p: 194°C–196°C, Rf .0.68(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.09 (d, J = 7.6 Hz, 1H, Ar-H), 7.95 (d, J = 8.7 Hz, 2H, Ar-Hp-OCH3 C6H4), 7.84 (t, J = 7.3 Hz, 1H, Ar-H), 7.59–7.47 (m, 3H, Ar-H), 7.44 (d, J = 7.6 Hz, 1H, Ar-H), 7.34 (d, J = 10.8 Hz, 2H, Ar-H),FIGURE 112D-Binding interactions of compound 9l within BRAFV600E (PDB ID: 5JRQ) active site showing H-bonds (as blue-arrows) and pi-pi interactions (asgreen-dotted lines) with Asp 594 and Phe 583, respectively.Frontiers in Chemistry frontiersin.org13Mohamed et al. 10.3389/fchem.2024.14476187.12 (d, J = 8.7 Hz, 2H, Ar-H p-OCH3 C6H4), 4.77 (s, 2H, S-CH2),3.85 (s, 3H, OCH3), 2.43 (s, 3H, CH3);13C NMR (100 MHz, δ ppmDMSO-d6): 176.66, 167.51, 161.72, 160.52, 155.52, 146.83, 139.28,135.37, 135.00, 130.85, 129.62, 129.43, 128.67, 126.58, 126.41,126.31, 125.89, 119.57, 118.28, 114.65, 55.37, 26.90, 20.75; Anal.Calc. (%) For C25H20N4O3S: C, 65.77; H, 4.42; N, 12.27; S, 7.02.Found: C, 65.62; H, 4.61; N, 12.49; S, 7.14.4.1.12 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-ethylquinazolin-4(3H)-one (9j)Yield: 0.17 g (79%), White solid, m.p: 114°C–116°C, Rf. 0.60(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.04 (d, J = 8.0 Hz, 1H, Ar-H), 7.97 (d, J = 6.4 Hz, 2H,Ar-H), 7.74 (t, J = 7.4 Hz, 1H, Ar-H), 7.57–7.54 (m, 3H, Ar-H), 7.42 (t,J = 7.4 Hz, 1H, Ar-H), 7.35 (d, J = 8.4 Hz, 1H, Ar-H), 4.94 (s, 2H,S-CH2), 4.11 (q, J = 7.0 Hz, 2H, N-CH2), 1.31 (t, J = 7.0 Hz, 3H, CH3);13C NMR (100MHz, δ ppmDMSO-d6): 177.21, 167.84, 160.13, 154.45,146.41, 134.73, 131.60, 129.27, 126.95, 126.38, 126.22, 126.00, 125.66,118.84, 26.70, 13.04; Anal. Calc. (%) For C19H16N4O2S: C, 62.62; H,4.43; N, 15.37; S, 8.80. Found: C, 62.89; H, 4.51; N, 15.62; S, 8.71.4.1.13 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-ethylquinazolin-4(3H)-one (9k)Yield: 0.20 g (84%), White solid, m.p: 118°C–120°C, Rf. 0.62(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.04 (dd, J = 7.9, 1.0 Hz, 1H, Ar-H), 7.98 (d, J = 8.6 Hz, 2H, Ar-Hp-Cl C6H4), 7.75–7.72 (m, 1H, Ar-H), 7.60 (d, J = 8.6 Hz, 2H, Ar-Hp-Cl C6H4), 7.44–7.41 (m, 1H, Ar-H), 7.35 (d, J = 8.1 Hz, 1H, Ar-H),4.94 (s, 2H, S-CH2), 4.11 (q, J = 7.0 Hz, 2H, N-CH2), 1.31 (t, J =7.0 Hz, 3H, CH3);13C NMR (100 MHz, δ ppm DMSO-d6): 177.49,167.05, 160.11, 154.42, 146.39, 136.35, 134.72, 129.43, 128.73,126.37, 126.22, 125.64, 124.82, 118.83, 39.58, 26.67, 13.03; Anal.Calc. (%) For C19H15ClN4O2S: C, 57.21; H, 3.79; N, 14.05; S, 8.04.Found: C, 57.49; H, 3.86; N, 14. 27; S, 8.12.4.1.14 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-ethylquinazolin-4(3H)-one (9l)Yield: 0.20 g (84%), White solid, m.p: 134°C–136°C, Rf. 0.57(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.04 (dd, J = 8.0, 1.3 Hz, 1H, Ar-H), 7.91 (d, J = 8.9 Hz,2H, Ar-H p-OCH3 C6H4), 7.76–7.71 (m, 1H, Ar-H), 7.44–7.40 (m, 1H,Ar-H), 7.37 (d, J = 7.6 Hz, 1H, Ar-H), 7.07 (d, J = 8.9 Hz, 2H, Ar-Hp-OCH3 C6H4), 4.91 (s, 2H, S-CH2), 4.11 (q, J = 7.1 Hz, 2H, N-CH2),3.81 (s, 3H, OCH3), 1.31 (t, J = 7.1 Hz, 3H, CH3);13C NMR (100MHz, δppm DMSO-d6): 176.74, 167.52, 161.71, 160.11, 154.42, 146.40, 134.69,128.63, 126.35, 126.9, 125.65, 118.83, 118.25, 114.63, 55.35, 39.56, 26.62,13.01; Anal. Calc. (%) For C20H18N4O3S: C, 60.90; H, 4.60; N, 14.20; S,8.13. Found: C, 61.13; H, 4.74; N, 14.37; S, 8.20.4.1.15 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-allylquinazolin-4(3H)-one (9m)Yield: 0.18 g (80%),White solid, m.p: 110°C–112°C, Rf. 0.61(Hexane:Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppm DMSO-d6): 8.04 (d,J = 7.8 Hz, 1H, Ar-H), 7.97 (d, J = 6.8 Hz, 2H, Ar-H), 7.75 (t, J = 7.5 Hz,1H, Ar-H), 7.62–7.49 (m, 3H, Ar-H), 7.43 (t, J = 7.5 Hz, 1H, Ar-H), 7.37(d, J = 8.1 Hz, 1H, Ar-H), 6.11–5.83 (m, 1H, =CH), 5.24 (d, J = 10.4 Hz,TABLE 4 Theoretical calculations of ADME properties of compounds (9a-o) using swissADME software.Cpd MW Lipinski Parametersa F MR Water SolubilitySilicos-IT classGI absorption BBBHBA HBD nrotb TPSA MLogP9a 412.46 5 0 5 99.11 3.6 0.55 116.84 Poor High No9b 446.91 5 0 5 99.11 4.08 0.55 121.85 Poor High No9c 442.49 6 0 6 108.34 3.28 0.55 123.33 Poor High No9d 426.49 5 0 5 99.11 3.81 0.55 121.81 Poor High No9e 460.94 5 0 5 99.11 4.29 0.55 126.82 Poor High No9f 456.52 6 0 6 108.34 3.49 0.55 128.3 Poor High No9g 426.49 5 0 5 99.11 3.81 0.55 121.81 Poor High No9h 460.94 5 0 5 99.11 4.29 0.55 126.82 Poor High No9i 456.52 6 0 6 108.34 3.49 0.55 128.3 Poor High No9j 364.42 5 0 5 99.11 2.92 0.55 101.57 Poor High No9k 398.87 5 0 5 99.11 3.15 0.55 106.58 Poor High No9l 394.45 6 0 6 108.34 2.35 0.55 108.06 Poor High No9m 376.43 5 0 6 99.11 3.08 0.55 105.9 Poor High No9n 410.88 5 0 6 99.11 3.3 0.55 110.91 Poor High No9o 406.46 6 0 7 108.34 2.5 0.55 112.4 Poor High NoMW, molecular weight; HBA, H-bond acceptor; HBD, H-bond donor; nrotb, no. of rotatable bonds; TPSA, topological polar surface area (Å2); MLogP, n-octanol/water distribution coefficient;F, Abbott bioavailability scores (0–1).aDrug lead-like character: MW ≤ 500, HBA ≤10, HBD ≤5, nrotb ≤10, TPSA ≤140, lipophilicity parameter MLogP ≤5; MR, 40–130; BBB, Blood-Brain Barrier.Frontiers in Chemistry frontiersin.org14Mohamed et al. 10.3389/fchem.2024.14476181H, =CH2), 5.15 (d, J = 17.3 Hz, 1H, =CH2), 4.92 (s, 2H, S-CH2), 4.72 (d,J = 4.0 Hz, 2H, N-CH2);13C NMR (100MHz, δ ppmDMSO-d6): 177.20,167.84, 160.19, 154.93, 146.44, 134.94, 131.64, 131.25, 129.30, 126.96,126.53, 126.37, 126.01, 125.74, 118.73, 117.60, 45.97, 26.80; Anal. Calc.(%) For C20H16N4O2S: C, 63.81; H, 4.28; N, 14.88; S, 8.52. Found: C,64.05; H, 4.42; N, 15.17; S, 8.64.4.1.16 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-allylquinazolin-4(3H)-one (9n)Yield: 0.21 g (85%), White solid, m.p: 115°C–117°C, Rf. 0.63(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.07 (dd, J = 8.0, 1.2 Hz, 1H, Ar-H), 7.99 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.80–7.74 (m, 1H, Ar-H), 7.63 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.49–7.43 (m, 1H, Ar-H), 7.39 (d, J = 8.2, 1H, Ar-H),6.04–5.90 (m, 1H, =CH), 5.26 (dd, J = 10.4, 1.2 Hz, 1H, =CH2), 5.17(dd, J = 17.2, 1.2 Hz, 1H, =CH2), 4.93 (s, 2H, S-CH2), 4.74 (d, J =5.1 Hz, 2H, N-CH2);13C NMR (100MHz, δ ppmDMSO-d6): 177.39,167.02, 160.10, 154.80, 146.37, 136.32, 134.81, 131.19, 129.38,128.68, 126.47, 126.27, 125.68, 124.80, 118.69, 117.61, 45.92,26.72; Anal. Calc. (%) For C20H15ClN4O2S: C, 58.46; H, 3.68; N,13.64; S, 7.80. Found: C, 58.70; H, 3.73; N, 13. 91; S, 7. 94.4.1.17 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-allylquinazolin 4(3H)-one (9o)Yield: 0.20 g (83%), White solid, m.p: 122°C–124°C, Rf. 0.59(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.07 (d, J = 7.6 Hz, 1H, Ar-H), 7.92 (d, J = 8.7 Hz, 2H, Ar-Hp-OCH3 C6H4), 7.78 (t, J = 7.2 Hz, 1H, Ar-H), 7.46 (t, J = 7.4 Hz, 1H,Ar-H), 7.40 (d, J = 8.4 Hz, 1H, Ar-H), 7.09 (d, J = 8.7 Hz, 2H, Ar-Hp-OCH3 C6H4), 6.01–5.91 (m, 1H, =CH), 5.26 (d, J = 10.4 Hz,1H, =CH2), 5.17 (d, J = 17.3 Hz, 1H, =CH2), 4.91 (s, 2H, S-CH2), 4.74(d, J = 4.7 Hz, 2H, 2H, N-CH2), 3.82 (s, 3H, OCH3);13C NMR(100MHz, δ ppmDMSO-d6): 176.73, 167.53, 161.73, 160.16, 154.89,146.43, 134.88, 131.24, 128.64, 126.51, 126.33, 125.74, 118.72,118.28, 117.58, 114.66, 55.37, 45.93, 26.70; Anal. Calc. (%) ForC21H18N4O3S: C, 62.05; H, 4.46; N, 13.78; S, 7.89. Found: C,62.29; H, 4.53; N, 14.02; S, 7.81.4.2 Biology4.2.1 Assay of cell viability of 9a-oThe human mammary gland epithelial (MCF-10A) normal cellline was used to test the viability of compounds 9a-o (Mahmoudet al., 2022; Mekheimer et al., 2022). Refer to SupplementaryAppendix A (Supplementary Material) for more details.4.2.2 Assay of antiproliferative actionThe MTT assay was used to investigate the antiproliferativeactivity of hybrids 9a-o versus four human cancer cell lines usingErlotinib as a control: colon cancer (HT-29) cell line, pancreaticcancer (Panc-1) cell line, lung cancer (A-549) cell line, and breastcancer (MCF-7) cell line (El-Sherief et al., 2019; Al-Wahaibi et al.,2022). See Supplementary Appendix A for more information.4.2.3 EGFR inhibitory assayThe EGFR-TK test (Abdel-Aziz et al., 2023) assessed theinhibitory potency of the most effective antiproliferativederivatives 9b, 9c, 9h, 9k, and 9l against EGFR. For more details,see Supplementary Appendix A.4.2.4 BRAFV600E inhibitory assayAn in vitro investigation assessed the anti-BRAFV600E activity of9b, 9c, 9h, 9k, and 9l (Youssif et al., 2022). Refer to SupplementaryAppendix A for more details.4.2.5 EGFRT790M inhibitory assayThe HTRF KinEASE-TK assay (Miles et al., 2020) was used toevaluate the inhibitory action of the most potent hybrids, 9b, 9c, and9h, against mutant-type EGFR (EGFRT790M). For more details, seeSupplementary Appendix A.4.2.6 Cell cycle analysis and apoptosis detectionCompound 9b was investigated for its effects on cell cycleprogression and apoptosis induction in A-549 cells. A lungcancer (A-549) cell line was treated for 24 h with an IC50concentration of 9b. The cell line was labeled with PI/AnnexinV, and flow cytometry was done with a BD FASC Caliber (El-Sherief et al., 2018). See Supplementary Appendix A formore details.4.3 Docking studyMolecular docking simulations of 15 derivatives (9a-o) wereperformed via Molecular Operating Environment (MOE®)software according to reported protocols (Abdel-Aziz et al.,2023) within the active site of EGFR tyrosine kinase domain(PDB ID: 1M17), mutant EGFR kinase domain T790M(EGFRT790M; PDB ID: 2JIU), and mutant BRAF kinase domain(BRAFV600E; PDB ID: 5JRQ) crystals structures downloaded fromRSCB protein data bank (https://www.rcsb.org/). For more details,see Supplementary Appendix A.4.4 Calculations of SwissADMEPharmacokinetics and drug-likeness prediction for all the newlysynthesized compounds was performed using the online toolSwissADME predictor software (http://www.swissadme.ch/) madeby the Swiss Institute of Bioinformatics.Data availability statementData will be available upon request from the authors.Author contributionsAM: Formal Analysis, Methodology, Software, Writing–originaldraft. OA-G: Investigation, Supervision, Validation, Visualization,Writing–original draft. YM: Methodology, Supervision,Writing–original draft. KD: Funding acquisition, Writing–originaldraft. SB: Investigation, Validation, Visualization, Writing–reviewand editing. BY: Conceptualization, Data curation, Formal Analysis,Frontiers in Chemistry frontiersin.org15Mohamed et al. 10.3389/fchem.2024.1447618Investigation, Methodology, Resources, Software, Supervision,Validation, Visualization, Writing–original draft, Writing–reviewand editing.FundingThe author(s) declare that financial support was received for theresearch, authorship, and/or publication of this article. This workwas funded by the Researchers Supporting Project Number(RSP2024R388) King Saud University, Riyadh, Saudi Arabia. Theauthors also acknowledge support from the KIT-Publication Fundof the Karlsruhe Institute of Technology.AcknowledgmentsThis work was funded by the Researchers Supporting ProjectNumber (RSP2024R388) King Saud University, Riyadh, SaudiArabia. The authors also acknowledge support from the KIT-Publication Fund of the Karlsruhe Institute of Technology.Conflict of interestThe authors declare that the research was conducted in theabsence of any commercial or financial relationships that could beconstrued as a potential conflict of interest.Publisher’s noteAll claims expressed in this article are solely those of the authors anddo not necessarily represent those of their affiliated organizations, orthose of the publisher, the editors and the reviewers. Any product thatmay be evaluated in this article, or claim that may be made by itsmanufacturer, is not guaranteed or endorsed by the publisher.Supplementary materialThe Supplementary Material for this article can be found onlineat: https://www.frontiersin.org/articles/10.3389/fchem.2024.1447618/full#supplementary-materialReferencesAbdel-Aziz, M., Beshr, E. A., Gomaa, H. A., Mostafa, Y. A., Youssif, B. G., Hayallah, A.M., et al. (2023). Discovery of new cyanopyridine/chalcone hybrids as dual inhibitors ofEGFR/BRAFV600E with promising antiproliferative properties. Arch. Pharm. 356 (4),2200464. doi:10.1002/ardp.202200464Alshammari, M. B., Aly, A. A., Youssif, B. G., Bräse, S., Ahmad, A., Brown, A. B., et al.(2022). Design and synthesis of new thiazolidinone/uracil derivatives asantiproliferative agents targeting EGFR and/or BRAFV600E. Front. Chem. 10,1076383. doi:10.3389/fchem.2022.1076383Al-Wahaibi, L. H., Abou-Zied, H. A., Beshr, E. A., Youssif, B. G., Hayallah, A. M., andAbdel-Aziz, M. (2023a). Design, synthesis, antiproliferative actions, and DFT studies ofnew bis–pyrazoline derivatives as dual EGFR/BRAFV600E inhibitors. Int. J. Mol. Sci. 24(10), 9104. doi:10.3390/ijms24109104Al-Wahaibi, L. H., Abou-Zied, H. A., Hisham, M., Beshr, E. A., Youssif, B. G., Bräse,S., et al. (2023b). Design, synthesis, and biological evaluation of novel 3-cyanopyridone/pyrazoline hybrids as potential apoptotic antiproliferative agents targeting EGFR/BRAFV600E inhibitory pathways. Molecules 28 (18), 6586. doi:10.3390/molecules28186586Al-Wahaibi, L. H., El-Sheref, E. M., Hammouda, M. M., and Youssif, B. G. (2023c).One-pot synthesis of 1-Thia-4-azaspiro [4.4/5] alkan-3-ones via schiff base: design,synthesis, and apoptotic antiproliferative properties of dual EGFR/BRAFV600Einhibitors. Pharmaceuticals 16 (3), 467. doi:10.3390/ph16030467Al-Wahaibi, L. H., El-Sheref, E. M., Hassan, A. A., Bräse, S., Nieger, M., Youssif, B. G.,et al. (2023d). Synthesis and structure determination of substituted thiazole derivativesas EGFR/BRAFV600E dual inhibitors endowed with antiproliferative activity.Pharmaceuticals 16 (7), 1014. doi:10.3390/ph16071014Al-Wahaibi, L. H., Gouda, A.M., Abou-Ghadir, O. F., Salem, O. I., Ali, A. T., Farghaly,H. S., et al. (2020). Design and synthesis of novel 2, 3-dihydropyrazino [1, 2-a] indole-1,4-dione derivatives as antiproliferative EGFR and BRAFV600E dual inhibitors. Bioorg.Chem. 104, 104260. doi:10.1016/j.bioorg.2020.104260Al-Wahaibi, L. H., Mahmoud, M. A., Mostafa, Y. A., Raslan, A. E., and Youssif, B. G.(2023e). Novel piperine-carboximidamide hybrids: design, synthesis, andantiproliferative activity via a multi-targeted inhibitory pathway. J. EnzymeInhibition Med. Chem. 38 (1), 376–386. doi:10.1080/14756366.2022.2151593Al-Wahaibi, L. H., Mostafa, Y. A., Abdelrahman, M. H., El-Bahrawy, A. H.,Trembleau, L., and Youssif, B. G. (2022). Synthesis and biological evaluation ofindole-2-carboxamides with potent apoptotic antiproliferative activity as EGFR/CDK2 dual inhibitors. Pharmaceuticals 15 (8), 1006. doi:10.3390/ph15081006Benassi, A., Doria, F., and Pirota, V. (2020). Groundbreaking anticancer activity of highlydiversified oxadiazole scaffolds. Int. J. Mol. Sci. 21 (22), 8692. doi:10.3390/ijms21228692Bhat, M. A., Tüzün, B., Alsaif, N. A., Khan, A. A., and Naglah, A. M. (2022). Synthesis,characterization, molecular modeling against EGFR target and ADME/T analysis ofnovel purine derivatives of sulfonamides. J. Mol. Struct. 1257, 132600. doi:10.1016/j.molstruc.2022.132600Boran, A. D., and Iyengar, R. (2010). Systems approaches to polypharmacology anddrug discovery. Curr. Opin. drug Discov. & Dev. 13 (3), 297–309.Daina, A., Michielin, O., and Zoete, V. (2017). SwissADME: a free web tool to evaluatepharmacokinetics, drug-likeness and medicinal chemistry friendliness of smallmolecules. Sci. Rep. 7 (1), 42717. doi:10.1038/srep42717Derbyshire, E. R., Mazitschek, R., and Clardy, J. (2012). Characterization ofPlasmodium liver stage inhibition by halofuginone. ChemMedChem 7 (5), 844–849.doi:10.1002/cmdc.201200045Do, B., Mace, M., and Rexwinkle, A. (2016). Idelalisib for treatment of B-cellmalignancies. Am. J. health-system Pharm. 73 (8), 547–555. doi:10.2146/ajhp150281El Mansouri, A.-E., Oubella, A., Maatallah, M., AitItto, M. Y., Zahouily, M., Morjani,H., et al. (2020). Design, synthesis, biological evaluation and molecular docking of newuracil analogs-1, 2, 4-oxadiazole hybrids as potential anticancer agents. Bioorg. & Med.Chem. Lett. 30 (19), 127438. doi:10.1016/j.bmcl.2020.127438El-Sherief, H. A., Youssif, B. G., Abdelazeem, A. H., Abdel-Aziz, M., and Abdel-Rahman, H. M. (2019). Design, synthesis and antiproliferative evaluation of novel 1, 2,4-triazole/schiff base hybrids with EGFR and B-RAF inhibitory activities. Anti-CancerAgents Med. Chem. Former. Curr. Med. Chemistry-Anti-Cancer Agents. 19 (5), 697–706.doi:10.2174/1871520619666181224115346El-Sherief, H. A., Youssif, B. G., Bukhari, S. N. A., Abdelazeem, A. H., Abdel-Aziz, M.,and Abdel-Rahman, H. M. (2018). Synthesis, anticancer activity and molecularmodeling studies of 1, 2, 4-triazole derivatives as EGFR inhibitors. Eur. J. Med.Chem. 156, 774–789. doi:10.1016/j.ejmech.2018.07.024Finn, R. S., Martin, M., Rugo, H. S., Jones, S., Im, S.-A., Gelmon, K., et al. (2016).Palbociclib and letrozole in advanced breast cancer. N. Engl. J. Med. 375 (20),1925–1936. doi:10.1056/nejmoa1607303Flaherty, K. T., Infante, J. R., Daud, A., Gonzalez, R., Kefford, R. F., Sosman, J., et al.(2012). Combined BRAF and MEK inhibition in melanoma with BRAFV600 mutations. N. Engl. J. Med. 367 (18), 1694–1703. doi:10.1056/nejmoa1210093Food, U., and Administration, D. (1997). Center for drug evaluation andresearch, guidance for industry: dissolution testing of immediate release solidoral dosage forms.Fu, R.-g., Sun, Y., Sheng, W.-b., and Liao, D.-f. (2017). Designing multi-targetedagents: an emerging anticancer drug discovery paradigm. Eur. J. Med. Chem. 136,195–211. doi:10.1016/j.ejmech.2017.05.016Hendawy, O. M. (2022). A comprehensive review of recent advances in the biologicalactivities of 1, 2, 4-oxadiazoles. Arch. Pharm. 355 (7), 2200045. doi:10.1002/ardp.202200045Hisham, M., Hassan, H. A., Gomaa, H. A., Youssif, B. G., Hayalah, A. M., and Abdel-Aziz, M. (2023). Design, synthesis, and antiproliferative activity of quinazolin-4-one/chalcone hybrids via the egfr inhibition pathway. Anti-Cancer Agents Med. Chem.Former. Curr. Med. Chemistry-Anti-Cancer Agents. 23 (17), 1932–1943. doi:10.2174/1871520623666230727104933Frontiers in Chemistry frontiersin.org16Mohamed et al. 10.3389/fchem.2024.1447618Hisham, M., Hassan, H. A., Gomaa, H. A., Youssif, B. G., Hayallah, A. M., and Abdel-Aziz, M. (2022). Structure-based design, synthesis and antiproliferative action of newquinazoline-4-one/chalcone hybrids as EGFR inhibitors. J. Mol. Struct. 1254, 132422.doi:10.1016/j.molstruc.2022.132422Keith, C. T., Borisy, A. A., and Stockwell, B. R. (2005). Multicomponent therapeuticsfor networked systems. Nat. Rev. Drug Discov. 4 (1), 71–78. doi:10.1038/nrd1609Kerru, N., Gummidi, L., Maddila, S., Gangu, K. K., and Jonnalagadda, S. B. (2020). Areview on recent advances in nitrogen-containing molecules and their biologicalapplications. Molecules 25 (8), 1909. doi:10.3390/molecules25081909Kitadai, N., and Maruyama, S. (2018). Origins of building blocks of life: a review.Geosci. Front. 9 (4), 1117–1153. doi:10.1016/j.gsf.2017.07.007Loboda, K. B., Valjavec, K., Štampar, M., Wolber, G., Žegura, B., Filipič, M., et al.(2020). Design and synthesis of 3, 5-substituted 1, 2, 4-oxadiazoles as catalytic inhibitorsof human DNA topoisomerase IIα. Bioorg. Chem. 99, 103828. doi:10.1016/j.bioorg.2020.103828Mahmoud, M. A., Mohammed, A. F., Salem, O. I., Gomaa, H. A., and Youssif, B. G.(2022). New 1, 3, 4-oxadiazoles linked with the 1, 2, 3-triazole moiety as antiproliferativeagents targeting the EGFR tyrosine kinase. Arch. Pharm. 355 (6), 2200009. doi:10.1002/ardp.202200009McLaughlin, N. P., Evans, P., and Pines, M. (2014). The chemistry and biology offebrifugine and halofuginone. Bioorg. & Med. Chem. 22 (7), 1993–2004. doi:10.1016/j.bmc.2014.02.040Meanwell, N. (2017). A synopsis of the properties and applications of heteroaromaticrings in medicinal chemistry. Adv. Heterocycl. Chem., 245–361. doi:10.1016/bs.aihch.2016.11.002Medina-Franco, J. L., Giulianotti, M. A.,Welmaker, G. S., andHoughten, R. A. (2013).Shifting from the single to the multitarget paradigm in drug discovery. Drug Discov.today 18 (9-10), 495–501. doi:10.1016/j.drudis.2013.01.008Mekheimer, R. A., Allam, S. M., Al-Sheikh, M. A., Moustafa, M. S., Al-Mousawi, S. M.,Mostafa, Y. A., et al. (2022). Discovery of new pyrimido [5, 4-c] quinolines as potentialantiproliferative agents with multitarget actions: rapid synthesis, docking, and ADMEstudies. Bioorg. Chem. 121, 105693. doi:10.1016/j.bioorg.2022.105693Miles, D., Chen, Y., Paprcka, S. L., Foley, C. N., Grange, R., Leleti, M. R., et al. (2020).Discovery and characterization of potent and selective AXL receptor tyrosine kinaseinhibitors for cancer therapy. Eur. J. Cancer 138, S38–S39. doi:10.1016/s0959-8049(20)31175-8Minin, D., Popkov, S., Pesochinskaya, K., and Aleksanov, D. (2023). Synthesis of [4-(3-aryl-1, 2, 4-oxadiazol-5-ylmethylthio)-2-methylphenoxy] acetic acids as new potential PPARagonists. Russ. Chem. Bull. 72 (7), 1648–1653. doi:10.1007/s11172-023-3944-5Mondaca, S., Lacouture, M., Hersch, J., and Yaeger, R. (2018). Balancing RAF, MEK,and EGFR inhibitor doses to achieve clinical responses and modulate toxicity in BRAFV600E colorectal cancer, 2. JCO precision oncol. doi:10.1200/PO.18.00088Moussa, G., Alaaeddine, R., Alaeddine, L. M., Nassra, R., Belal, A. S., Ismail, A., et al.(2018). Novel click modifiable thioquinazolinones as anti-inflammatory agents: design,synthesis, biological evaluation and docking study. Eur. J. Med. Chem. 144, 635–650.doi:10.1016/j.ejmech.2017.12.065Notarangelo, T., Sisinni, L., Condelli, V., and Landriscina, M. (2017). Dual EGFR andBRAF blockade overcomes resistance to vemurafenib in BRAF mutated thyroidcarcinoma cells. Cancer cell Int. 17 (1), 86–89. doi:10.1186/s12935-017-0457-zOkaniwa, M., Hirose, M., Imada, T., Ohashi, T., Hayashi, Y., Miyazaki, T., et al.(2012). Design and synthesis of novel DFG-out RAF/vascular endothelial growth factorreceptor 2 (VEGFR2) inhibitors. 1. Exploration of [5, 6]-fused bicyclic scaffolds. J. Med.Chem. 55 (7), 3452–3478. doi:10.1021/jm300126xPadwa, A., and Bur, S. K. (2007). The domino way to heterocycles. Tetrahedron 63(25), 5341–5378. doi:10.1016/j.tet.2007.03.158Paraiso, K., Fedorenko, I., Cantini, L., Munko, A., Hall, M., Sondak, V., et al. (2010).Recovery of phospho-ERK activity allows melanoma cells to escape from BRAFinhibitor therapy. Br. J. cancer 102 (12), 1724–1730. doi:10.1038/sj.bjc.6605714Purcell, J. W., Davis, J., Reddy, M., Martin, S., Samayoa, K., Vo, H., et al. (2010).Activity of the kinesin spindle protein inhibitor ispinesib (SB-715992) in models ofbreast cancer. Clin. Cancer Res. 16 (2), 566–576. doi:10.1158/1078-0432.ccr-09-1498Raghavendra, N. M., Pingili, D., Kadasi, S., Mettu, A., and Prasad, S. (2018). Dual ormulti-targeting inhibitors: the next generation anticancer agents. Eur. J. Med. Chem.143, 1277–1300. doi:10.1016/j.ejmech.2017.10.021Robert, C., Karaszewska, B., Schachter, J., Rutkowski, P., Mackiewicz, A., Stroiakovski,D., et al. (2015). Improved overall survival in melanoma with combined dabrafenib andtrametinib. N. Engl. J. Med. 372 (1), 30–39. doi:10.1056/nejmoa1412690Sharma, V., Kamal, R., and Kumar, V. (2017). Heterocyclic analogues as kinaseinhibitors: a focus review. Curr. Top. Med. Chem. 17 (22), 2482–2494. doi:10.2174/1568026617666170307113744Umar, A. B., Uzairu, A., Shallangwa, G. A., and Uba, S. (2020). QSAR modelling andmolecular docking studies for anti-cancer compounds against melanoma cell line SK-MEL-2. Heliyon 6 (3), e03640. doi:10.1016/j.heliyon.2020.e03640Unadkat, V., Rohit, S., Parikh, P., Sanna, V., and Singh, S. (2021). Rational design-aided discovery of novel 1, 2, 4-oxadiazole derivatives as potential EGFR inhibitors.Bioorg. Chem. 114, 105124. doi:10.1016/j.bioorg.2021.105124Wahid, M., Jawed, A., Mandal, R. K., Dar, S. A., Akhter, N., Somvanshi, P., et al.(2018). Recent developments and obstacles in the treatment of melanoma with BRAFand MEK inhibitors. Crit. Rev. oncology/hematology 125, 84–88. doi:10.1016/j.critrevonc.2018.03.005Yan, G., Zhang, Y., and Wang, J. (2017). Recent advances in the synthesis of aryl nitrilecompounds. Adv. Synthesis & Catal. 359 (23), 4068–4105. doi:10.1002/adsc.201700875Youssif, B. G., Gouda, A. M., Moustafa, A. H., Abdelhamid, A. A., Gomaa, H. A.,Kamal, I., et al. (2022). Design and synthesis of new triarylimidazole derivatives as dualinhibitors of BRAFV600E/p38α with potential antiproliferative activity. J. Mol. Struct.1253, 132218. doi:10.1016/j.molstruc.2021.132218Youssif, B. G., Mohamed, M. F., Al-Sanea, M. M., Moustafa, A. H., Abdelhamid,A. A., and Gomaa, H. A. (2019). Novel aryl carboximidamide and 3-aryl-1, 2, 4-oxadiazole analogues of naproxen as dual selective COX-2/15-LOX inhibitors:design, synthesis and docking studies. Bioorg. Chem. 85, 577–584. doi:10.1016/j.bioorg.2019.02.043Zhang, Q., Diao, Y., Wang, F., Fu, Y., Tang, F., You, Q., et al. (2013). Design anddiscovery of 4-anilinoquinazoline ureas as multikinase inhibitors targeting BRAF,VEGFR-2 and EGFR. MedChemComm 4 (6), 979–986. doi:10.1039/c3md00096fZhou, J., Jiang, X., He, S., Jiang, H., Feng, F., Liu, W., et al. (2019). Rational design ofmultitarget-directed ligands: strategies and emerging paradigms. J. Med. Chem. 62 (20),8881–8914. doi:10.1021/acs.jmedchem.9b00017Frontiers in Chemistry frontiersin.org17Mohamed et al. 10.3389/fchem.2024.1447618', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explai"}}},{"rowIdx":90,"cells":{"Researcher Name":{"kind":"string","value":"Dongbo Wu"},"ORCID":{"kind":"string","value":"0000-0003-4500-047X"},"Researcher Query Keyword":{"kind":"string","value":"Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":91,"cells":{"Researcher Name":{"kind":"string","value":"Jin-Bu Xu"},"ORCID":{"kind":"string","value":"0009-0005-4707-5186"},"Researcher Query Keyword":{"kind":"string","value":"Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":92,"cells":{"Researcher Name":{"kind":"string","value":"Lan Zhang"},"ORCID":{"kind":"string","value":"0000-0002-7572-1035"},"Researcher Query Keyword":{"kind":"string","value":"Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives': 'Title: Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives\\\\nDesign and synthesis of new1,2,4-oxadiazole/quinazoline-4-one hybrids withantiproliferative activity asmultitargeted inhibitorsAmira M. Mohamed1, Ola M. F. Abou-Ghadir1, Yaser A. Mostafa1,Kholood A. Dahlous2, Stefan Bräse3* and Bahaa G. M. Youssif1*1Pharmaceutical Organic Chemistry Department, Faculty of Pharmacy, Assiut University, Assiut, Egypt,2Department of Chemistry, College of Science, King Saud University, Riyadh, Saudi Arabia, 3Institute ofBiological and Chemical Systems, IBCS-FMS, Karlsruhe Institute of Technology, Karlsruhe, GermanyIntroduction: The combination of BRAF and tyrosine kinase (TK) inhibitors hasbeen demonstrated to be highly effective in inhibiting tumor development and isan approach for overcoming resistance in clinical trials. Accordingly, a novelseries of 1,2,4-oxadiazole/quinazoline-4-one hybrids was developed asantiproliferative multitargeted inhibitors.Methods: The structures of the newly synthesized compounds 9a-o werevalidated using IR, NMR, MS, and elemental techniques. 9a–o were tested asantiproliferative agents.Results and Discussion: The results showed that the majority of the testedcompounds showed significant antiproliferative action with 9b, 9c, 9h, 9k, and9l being the most potent. Compounds 9b, 9c, 9h, 9k, and 9l were tested as EGFRand BRAFV600E inhibitors. These in vitro tests revealed that compounds 9b, 9c, and9h are strong antiproliferative agents that may act as dual EGFR/BRAFV600Einhibitors. 9b, 9c, and 9h were further investigated for their inhibitory effecton mutant EGFR (EGFRT790M), and the results showed that the tested compoundshad considerable inhibitory action. Cell cycle study and apoptosis detectiondemonstrated that compound 9b exhibits cell cycle arrest at the G2/Mtransition. Molecular docking simulations reveal the binding mechanism of themost active antiproliferative agents.KEYWORDSquinazolinone, oxadiazole, kinases, apoptosis, antiproliferative, EGFR, BRAF1 IntroductionDrug developers have spent decades generating selective medicines for specific targets(Medina-Franco et al., 2013; Zhou et al., 2019). Despite the effectiveness of many single-target selective medications, the advancement of multifactorial disorders such as cancer andneurological diseases included many signaling pathways (Fu et al., 2017; Raghavendra et al.,2018). As a result, there is a growing interest in developing treatments that address manytargets at the same time.OPEN ACCESSEDITED BYXuekui Xia,Biology Institute of Shandong Academy ofSciences, ChinaREVIEWED BYChao Liu,Shandong University, ChinaRavikumar Jimmidi,Baylor College of Medicine, United States*CORRESPONDENCEBahaa G. M. Youssif,bgyoussif2@gmail.com,bahaa.youssif@pharm.aun.edu.egStefan Bräse,braese@kit.eduRECEIVED 11 June 2024ACCEPTED 05 August 2024PUBLISHED 30 August 2024CITATIONMohamed AM, Abou-Ghadir OMF, Mostafa YA,Dahlous KA, Bräse S and Youssif BGM (2024)Design and synthesis of new 1,2,4-oxadiazole/quinazoline-4-one hybrids withantiproliferative activity asmultitargeted inhibitors.Front. Chem. 12:1447618.doi: 10.3389/fchem.2024.1447618COPYRIGHT© 2024 Mohamed, Abou-Ghadir, Mostafa,Dahlous, Bräse and Youssif. This is an open-access article distributed under the terms of theCreative Commons Attribution License (CC BY).The use, distribution or reproduction in otherforums is permitted, provided the originalauthor(s) and the copyright owner(s) arecredited and that the original publication in thisjournal is cited, in accordance with acceptedacademic practice. No use, distribution orreproduction is permitted which does notcomply with these terms.Frontiers in Chemistry frontiersin.org01TYPE Original ResearchPUBLISHED 30 August 2024DOI 10.3389/fchem.2024.1447618There are currently two opposing methodologies for designingmulti-targeting medicines. The first technique involves establishingan additive or synergistic effect of various medications operating onseparate targets through combination drug therapy. Preclinicalevidence of enhanced apoptosis and delayed resistance to BRAF(Rapidly Accelerated Fibrosarcoma, B-family) inhibitors (Paraisoet al., 2010; Flaherty et al., 2012), for example, prompted the FDA toapprove a combination of dabrafenib (BRAF inhibitor) andtrametinib (MEK inhibitor) for the treatment of metastaticmelanoma with BRAF mutations (Robert et al., 2015; Wahidet al., 2018). The use of palbociclib and letrozole in the treatmentof advanced breast cancer is another example of successfulcombination therapy (Finn et al., 2016).The second approach is designing and generating multiple-targeting medicines that synergistically block numerouscarcinogenic pathways (Keith et al., 2005; Boran and Iyengar,2010). The method of multi-targeting therapies is finding a singleagent that can operate on two or more targets simultaneously.Cabozantinib, also known as cabometyx, was approved by theFDA as a small molecule dual-targeting inhibitor of the tyrosinekinases c-Met (mesenchymal-epithelial transition factor) andVEGFR-2 (Vascular Endothelial Growth Factor Receptor) andhas been demonstrated to suppress tumor growth, metastasis,and angiogenesis (Food and Administration, 1997).On the other hand, the combination of BRAF and tyrosinekinase (TK) inhibitors has been demonstrated to be highly effectivein inhibiting tumor development and is an approach for overcomingresistance in clinical trials. Vemurafenib (BRAFV600E inhibitor)resistance in thyroid cancer can be addressed by combining itwith EGFR (Epidermal Growth Factor Receptor) inhibitors(Notarangelo et al., 2017). This combination has also shownpromising results in BRAFV600E colorectal cancer (Mondaca et al.,2018). In addition, various compounds have been discovered in vitrothat include the key pharmacophoric groups required to inhibittyrosine kinases, such as EGFR/VEGFR-2 and BRAF (Okaniwaet al., 2012; Zhang et al., 2013). Compound I (Figure 1) inhibitedwild-type BRAF and EGFR with IC50 values in the nanomolar range.Additionally, imidazo [1,2-b]pyridazine II inhibited BRAF andVEGFR-2.Heterocyclic moieties form the largest and most varied class oforganic molecules. In medicinal chemistry, compounds containingheterocyclic nuclei have gained great interest because of their diversetherapeutic actions (Padwa and Bur, 2007). Heterocyclics play acrucial role in the breakdown of all living things and participate invarious biochemical processes necessary for life (Kitadai andMaruyama, 2018). The heteroaromatic framework resemblesbiologically active compounds such as nucleic acids, hormones,and neurotransmitters (Meanwell, 2017). As a result, thesemoieties could be used to design safer medications. Heterocyclesare often found in nature and have been exploited to develop anti-cancer drugs that target many sites and disrupt cancer growthpathways (Sharma et al., 2017). Heterocyclic rings can bemodified with various substituents to cover various chemicals,making them ideal for designing anti-cancer drugs.Nitrogen-containing heterocyclic chemicals significantly affectabout 75% of FDA-approved anti-cancer drugs (Kerru et al., 2020).Quinazolinone, a bicyclic system composed of benzene andpyrimidinone rings, is one of the most common nitrogen-containing heterocycles in medicinal chemistry, found in variouscompounds with diverse biological activity. Idelalisib III (Do et al.,2016), Ispinesib IV (Purcell et al., 2010), and Halofuginone V(Figure 2) (Derbyshire et al., 2012; McLaughlin et al., 2014) areexamples of recently approved or marketed medications with anti-cancer properties.Depending on the position of the keto or oxo group, threedistinct forms are possible: quinazolin-2(1H)-one VI, quinazoline-2,4-(1H,3H)-di-one VII, and quinazolin-4(3H)-one VIII (Figure 3).Among these, quinazolin-4-one VIII is the most commonly usedscaffold in synthetic processes or as a structural component ofnatural compounds (Sharma et al., 2017). This last scaffold isadaptable, allowing up to six potential substitutes in positions 2,3, 5, 6, 7, and 8.In recent publications (Hisham et al., 2022; Hisham et al., 2023),we present the design and synthesis of a new series of quinazoline-4-one/chalcone hybrids that function as dual inhibitors of EGFR andBRAFV600E with antiproliferative activity. The target compoundswere tested in vitro against various cancer cell lines and the EGFRand BRAFV600E enzymes. Compound IX (Figure 4) was the mostpotent derivative, with a GI50 of 1.16 µM, compared to the referencedrug Doxorubicin (GI50 = 1.14 µM). Compound IX showedsignificant inhibitory activity against EGFR and BRAFV600E, withIC50 values of 0.11 µM and 0.65 µM, respectively. Moreover,apoptosis assay results revealed that compound IX enhanced thelevel of active caspase-3, 8, and 9 with significant induction ofcytochrome c and Bax levels and downregulation of the anti-apoptotic Bcl-2 levels.FIGURE 1Structures of compounds I and II.Frontiers in Chemistry frontiersin.org02Mohamed et al. 10.3389/fchem.2024.1447618On the other hand, literature reviews reveal that 1,2,4-oxadiazoles have statistical significance in bioorganic andmedicinal chemistry. They are recognized for their diversepharmacological characteristics (Benassi et al., 2020; El Mansouriet al., 2020; Loboda et al., 2020). The 1,2,4-oxadiazole exhibitsbioisosteric equivalence with ester and amide moieties. Whenunstable conditions (e.g., hydrolysis) are identified, 1,2,4-oxadiazole is a highly effective alternative (Hendawy, 2022). Thesubstantial biological impact of 1,2,4-oxadiazole derivatives oncancer cells can be attributed to various mechanisms of action.For example, we developed and synthesized novel 1,2,4-oxadiazole-based derivatives linked to a triaryl-imidazole moiety, withcompound X (Figure 4) being the most potent (Youssif et al.,2022). In vitro studies assessed the antiproliferative effects ofrecently identified compounds inhibiting p38α and BRAFV600E.These compounds showed effective antiproliferative and kinaseinhibition.Another set of 1,2,4-oxadiazole-based compounds (compoundXI, Figure 4) were designed, synthesized, and tested forantiproliferative properties against EGFR-TK. The experimentshowed promising antiproliferative effects against cancer celllines, with low micromolar IC50 values against EGFR, comparedto the reference doxorubicin (Unadkat et al., 2021).1.1 Rational designConsistent with prior findings and continuing our efforts todevelop dual or multitargeted antiproliferative agents (Al-Wahaibiet al., 2020; Alshammari et al., 2022; Al-Wahaibi et al., 2023a; Abdel-Aziz et al., 2023; Al-Wahaibi et al., 2023b; Al-Wahaibi et al., 2023c;Al-Wahaibi et al., 2023d; Al-Wahaibi et al., 2023e), this study’sstrategy was to design and synthesize new antiproliferative agentsbased on quinazoline-4-one/1,2,4-oxadiazole hybrids (Figure 5) toobtain new anti-tumor agents with synergistic activity.The substitutions on the nitrogen atom of the quinazolinemoiety were changed between alkyl (methyl, ethyl, and allyl) andaryl (phenyl or tolyl) moieties to examine the impact of rigidity onthe antiproliferative activity of these compounds. In addition,different substituents, such as a chlorine atom (electronFIGURE 2Examples of approved or commercialized anti-cancer medicines with the quinazoline-4-one scaffold.FIGURE 3Different forms of quinazolinones.Frontiers in Chemistry frontiersin.org03Mohamed et al. 10.3389/fchem.2024.1447618withdrawing group) or a methoxy group (electron donating group),were used to investigate the different electronic impacts of thesesubstituents on 9a-o’s antiproliferative activity.All of the novel analogs were examined for cell viability effectagainst a normal human mammary gland epithelial (MCF-10A) cellline as well as for antiproliferative activity against four human cancercell lines: colon (HT-29), pancreatic (Panc-1), lung (A-549), andbreast (MCF-7). Furthermore, the compounds with the highestantiproliferative activity were investigated in vitro as multi-targeting inhibitors of EGFR, EGFRT790M, and BRAFV600Eenzymes. The study was expanded to include one of the mostactive derivatives, 9b, as a representative agent to evaluate itsFIGURE 4Structures of quinazoline-4-one and 1,2,4-oxadiazole-based derivatives as anticancer agents.FIGURE 5Structures of new targets 9a-o.Frontiers in Chemistry frontiersin.org04Mohamed et al. 10.3389/fchem.2024.1447618mechanistic effects on the cell cycle and induction of apoptosis.Finally, docking studies were performed on the most activecompounds against the selected enzymes to explain their in vitroresults. Furthermore, the ADME analyses were performed toinvestigate their pharmacokinetic features.2 Results and discussion2.1 ChemistryScheme 1 summarizes the synthetic pathways of the new targetcompounds 9a-o. Anthranilic acid (1) was refluxed in ethanol withisothiocyanate derivatives 2a-e for 8 h. After the reaction wascompleted (as determined by TLC), the resulting whiteprecipitate was collected by filtration and recrystallized from anethanol/dioxane mixture (1:1) to give the corresponding quinazolinederivatives 3a-e in 90%–95% yields (Moussa et al., 2018). On theother hand, compounds 6a-c, amidoxime derivatives, weresynthesized in 50%–60% yields over two steps, Scheme 2. Thefirst step involved reacting the corresponding aldehydes 4a-c with28% aqueous ammonia and iodine in THF for 2–3 h to yield theSCHEME 1Synthesis of the new target compounds 9a-oSCHEME 2Structure Activity Relationship (SAR) analysis of new targets 9a-o.Frontiers in Chemistry frontiersin.org05Mohamed et al. 10.3389/fchem.2024.1447618corresponding aryl nitrile derivatives 5a-c in 76%–80% (Yan et al.,2017). The second step was a 12- to 18-h methanol reflux ofcompounds 5a-c with hydroxylamine chloride and sodiumcarbonate (Youssif et al., 2019). Compounds 6a-c were reactedwith chloroacetyl chloride in dry acetone to yieldbenzimidamides (7a-c), which were cyclized by refluxing intoluene to the corresponding 3-aryl-5-(chloromethyl)-1,2,4-oxadiazole derivatives 8a-c as a yellow oil. Compounds 8a-c werepurified using column chromatography with hexane: ethyl acetate(9:1) as an eluent (Minin et al., 2023). For example, the 1H NMRspectrum of compound 8b confirmed the disappearance of twoprotons from the NH2 group of the corresponding amidoxime 7b.Moreover, the spectrum displayed a singlet signal corresponding tothe methylene protons (Cl-CH2) at δ 4.74. The spectra also revealeda characteristic pair of doublets in the aromatic region for 4-ClC6H4at δ 8.01 and 7.46.Reagents and Conditions: a) Triethylamine, ethanol, Reflux 8 h;b) ammonia (28%), I2, THF, Stirring 1h; c) NH2OHHCl, Na2CO3,THF, Reflux 18 h; d) Chloroacetyl Chloride, K2CO3, Dry acetone,Stirring 24hrs; e) Toluene, Reflux 10 h; f) K2CO3, KI, DMF,Stirring 24 h.Finally, the target novel compounds, 9a-o, were synthesizedin high yields by coupling compounds 3a-e with thecorresponding 1,2,4-oxadiazoles 8a-c in DMF using anhydrousK2CO3 and KI and stirring for 24 h at room temperature. 9a-owere purified via ethanol recrystallization. The structures of 9a-owere elucidated using 1H NMR, 13C NMR, and elementalmicroanalyses. The 1H NMR spectra of compound 9l, as anexample, confirmed the presence of ethyl group characteristicsignals in the form of triplet at δ 1.31 (t, J = 7.1 Hz, 3H,N-CH2CH3) and quartet at δ 4.11 (q, J = 7.1 Hz, 2H, N-CH2).The spectrum also revealed two distinct singlet signals: at δ 3.81(s, 3H, OCH3) and at δ 4.91 (s, 2H, S-CH2). Additionally, thespectrum revealed a pair of doublets of the aromatic ring’s para disubstitution pattern and extra signals for the aromatic protons inthe quinazoline moiety. The 13C NMR spectrum of 9l indicatedthe presence of ethyl group characteristic signals at δ 39.56 and δ13.01, methylene group at δ 26.62, and methoxy group at δ 55.TABLE 1 Cell viability percent and antiproliferative activity (IC50 values) of compounds 9a-o.Comp Cell viability % R1 R2 Antiproliferative activity IC50 ± SEM (nM)A-549 MCF-7 Panc-1 HT-29 Average (GI50)9a 90 Phenyl H 46 ± 4 50 ± 4 48 ± 4 48 ± 4 489b 89 Phenyl Cl 22 ± 2 26 ± 2 24 ± 2 24 ± 2 249c 91 Phenyl OMe 24 ± 2 28 ± 3 26 ± 2 25 ± 2 269d 90 p-Tolyl H 54 ± 4 58 ± 5 54 ± 5 55 ± 5 559e 91 p-Tolyl Cl 40 ± 3 44 ± 4 42 ± 4 42 ± 4 429f 92 p-Tolyl OMe 50 ± 4 53 ± 5 52 ± 5 52 ± 5 529g 90 m-Tolyl H 49 ± 4 52 ± 5 50 ± 5 50 ± 5 509h 89 m-Tolyl Cl 28 ± 2 31 ± 3 29 ± 2 30 ± 3 309i 91 m-Tolyl OMe 65 ± 6 69 ± 6 66 ± 6 68 ± 6 679j 90 Ethyl H 69 ± 6 76 ± 7 68 ± 6 68 ± 6 709k 93 Ethyl Cl 32 ± 3 35 ± 3 34 ± 3 34 ± 3 349l 90 Ethyl OMe 36 ± 3 40 ± 4 38 ± 3 37 ± 3 389m 92 Allyl H 62 ± 6 65 ± 6 64 ± 6 62 ± 6 639n 91 Allyl Cl 43 ± 4 47 ± 4 45 ± 4 44 ± 4 459o 89 Allyl OMe 56 ± 4 59 ± 5 57 ± 5 57 ± 5 57Erlotinib ND — — 30 ± 3 40 ± 3 30 ± 3 30 ± 3 33ND: not determined.Frontiers in Chemistry frontiersin.org06Mohamed et al. 10.3389/fchem.2024.1447618Elemental microanalysis of 9l confirmed that the calculated data(%) were C, 60.90; H, 4.60; N, 14.20; S, 8.13, while the found data(%) were C, 61.13; H, 4.74; N, 14.37; S, 8.20.2.2 Biology2.2.1 Assay of cell viability effectThe human mammary gland epithelial (MCF-10A) normal cellline was used to test the viability of novel compounds 9a-o(Mahmoud et al., 2022; Mekheimer et al., 2022). After 4 days ofincubation on MCF-10A cells, the cell viability of compounds 9a-owas determined using the MTT test. Table 1 demonstrates that noneof the compounds examined were cytotoxic, and all hybrids showedmore than 89% cell viability at 50 µM.2.2.2 Assay of antiproliferative actionThe MTT assay was used to investigate the antiproliferativeactivity of hybrids 9a-o versus four human cancer cell lines usingErlotinib as a control: colon cancer (HT-29) cell line, pancreaticcancer (Panc-1) cell line, lung cancer (A-549) cell line, and breastcancer (MCF-7) cell line (El-Sherief et al., 2019; Al-Wahaibi et al.,2022). Table 1 displays the median inhibitory concentration (IC50)and GI50 (average IC50) against the four cancer cell lines.In general, the hybrids 9a-o had significant antiproliferative actionwith GI50 values ranging from 24 nM to 70 nM versus the four cancercell lines evaluated, compared to Erlotinib, which had a GI50 value of33 nM. Compounds 9b, 9c, 9h, 9k, and 9l were the most potent fivederivatives, with GI50 values of 24, 26, 30, 34, and 38 nM,making 9b, 9c,and 9hmore potent than Erlotinib (GI50 = 33 nM). Out of all the newlysynthesized hybrids 9a-o, compound 9b (R1 = phenyl, R2 = Cl) had thehighest potency, with a GI50 value of 24 nM, which was 1.4 times morepotent than the reference Erlotinib (GI50 = 33 nM).The type of the aryl/alkyl moieties at position 3 of thequinazoline moiety appears to be critical for 9a-o hybridsantiproliferative activity. The GI50 value of compound 9h (R1 =m-tolyl, R2 = Cl) was 30 nM, less potent than compound 9b but stillmore potent than the reference erlotinib (GI50 = 33 nM).Moreover, Compounds 9e (R1 = p-tolyl, R2 = Cl), 9k (R1 = ethyl,R2 = Cl), and 9n (R1 = allyl, R2 = Cl) demonstrated GI50 values of 42,34, and 45 nM, respectively, being less potent than compounds 9b,9h, and even Erlotinib. These results indicated the importance of thequinazoline moiety position three substitution pattern onantiproliferative activity, with activity rising in the followingorder: phenyl > m-tolyl > p-tolyl > ethyl > allyl.Compound 9c (R1 = phenyl, R2 = OMe) rated second in activityagainst the four cancer cell lines, with a GI50 value of 26 nM, slightlyless effective than 9b but still more potent than Erlotinib (GI50 =33 nM). The unsubstituted phenyl derivative, 9a (R1 = phenyl, R2 =H), was less potent than 9b and 9c, indicating that the substitutionpattern at the fourth position of the phenyl group in the 1,2,4-oxadiazole moiety affects the antiproliferative activity of thesehybrids, with activity increasing in the order Cl > OMe > H.Regardless of the nature of the substitution pattern at position3 of the quinazoline moiety, the same rule (Cl > OMe > H inactivity) applies to other derivatives.2.2.3 EGFR inhibitory assayThe EGFR-TK test (Abdel-Aziz et al., 2023) was used to assessthe inhibitory potency of the most effective antiproliferativederivatives 9b, 9c, 9h, 9k, and 9l against EGFR, and the resultsare shown in Table 2. This assay’s results are in line with theantiproliferative assay’s, which found that compounds 9b (R1 =phenyl, R2 = Cl), 9c (R1 = phenyl, R2 = OMe), and 9h (R = m-tolyl,R2 = Cl), the most effective antiproliferative hybrids, were the mostpotent derivatives of EGFR inhibitors, with IC50 values of 57 ± 4 nM,64 ± 5 nM, and 72 ± 5 nM, respectively, surpassing the referencedrug Erlotinib (IC50 = 80 ± 5). Compounds 9k (R1 = ethyl, R2 = Cl)and 9l (R1 = ethyl, R2 = OMe) demonstrated significant anti-EGFRactivity, with IC50 values of 84 ± 6 nM and 91 ± 07 nM, respectively,which were less potent than Erlotinib. These findings show thatcompounds 9b, 9c, and 9h have significant EGFR inhibitory actionand could be used as antiproliferative agents.2.2.4 BRAFV600E inhibitory assayAn in vitro investigation assessed the anti-BRAFV600E activity of9b, 9c, 9h, 9k, and 9l (Youssif et al., 2022). The enzyme assaydemonstrated that the five hybrids examined substantially inhibitedBRAFV600E, with IC50 values ranging from 48 to 70 nM, Table 2. In allcases, the IC50 of the examined compounds is greater than that of thereference Vemurafenib (IC50 = 30). Compounds 9b, 9c, and 9hdemonstrated the most effective inhibitory activity againstTABLE 2 IC50 values of compounds 9b, 9c, 9h, 9k, and 9l against EGFR, BRAFV600E and EGFRT790M.Compound EGFR inhibitionIC50 ± SEM (nM)BRAFV600E inhibition IC50 ± SEM (nM) EGFRT790M inhibition IC50 ± SEM (nM)9b 57 ± 4 48 ± 4 10 ± 19c 64 ± 5 51 ± 5 11 ± 29h 72 ± 5 57 ± 5 15 ± 29k 84 ± 6 62 ± 5 ND9l 91 ± 7 70 ± 5 NDErlotinib 80 ± 5 60 ± 5 NDVemurafenib ND 30 ± 3 NDOsimertinib ND ND 8 ± 1ND: not determined.Frontiers in Chemistry frontiersin.org07Mohamed et al. 10.3389/fchem.2024.1447618BRAFV600E (IC50 = 48, 51, and 57 nM, respectively) and werediscovered to be potent inhibitors of cancer cell growth (GI50 =24, 26, and 30 nM, respectively). As a result, compounds 9b, 9c, and9h are effective antiproliferative agents that function as dual EGFR/BRAFV600E inhibitors.2.2.5 EGFRT790M inhibitory assayThe HTRF KinEASE-TK assay (Miles et al., 2020) was used toevaluate the inhibitory action of the most potent hybrids, 9b, 9c, and9h, against mutant-type EGFR (EGFRT790M). As demonstrated inTable 2, compounds 9b, 9c, and 9h displayed excellent inhibitoryeffect against EGFRT790M, with IC50 values of 10 ± 1, 11 ± 1, and 15 ±1 nM, respectively, being equivalent to the reference Osimertinib(IC50 = 8 ± 1 nM), which may explain their robust antiproliferativeactivity. These findings suggested that phenyl and m-tolylsubstitutions in the quinazoline moiety’s third position, as well aschlorine atom and methoxy substitutions in the para-position of thephenyl group in the 1,2,4-oxadiazole moiety, are required for theinhibitory impact on EGFR, BRAFV600E, and EGFRT790M.1. The type of the aryl/alkyl moieties at position 3 of thequinazoline moiety appears to be critical for 9a-o hybridsantiproliferative activity, with activity increasing in thefollowing order: phenyl > m-tolyl > p-tolyl > ethyl > allyl.2. The substitution pattern at the fourth position of the phenylgroup in the 1,2,4-oxadiazole moiety affects theantiproliferative activity of these hybrids as well as EGFR,BRAFV600E, and EGFRT790M inhibition, with activity increasingin the order Cl > OMe > H.3. Regardless of the nature of the substitution pattern at position3 of the quinazoline moiety, the same rule (Cl > OMe > H inactivity) applies to other derivatives.2.2.6 Cell cycle analysis and apoptosis assays2.2.6.1 Cell cycle analysisCompound 9b was investigated for its effects on cell cycleprogression and apoptosis induction in A-549 cells. A lungcancer (A-549) cell line was treated for 24 h with an IC50concentration of 9b. The cell line was labeled with PI/Annexin V,and flow cytometry was done with a BD FASC Caliber (El-Sheriefet al., 2018). The results (Figure 6) showed that A-549 treated withcompound 9b had a significant percentage of cell accumulation(29%) in the G2/M phase after 24 h of incubation, indicating cellcycle arrest at the G2/M transition.2.2.6.2 Apoptosis induction assayTo assess 9b′s potential to induce apoptosis, A-549 cells werelabeled with Annexin V/PI, grown for 24 h, and evaluated.Examining early and late apoptosis demonstrated that compound9b could produce high levels of apoptosis, with a necrosis percentage6.43 (Figures 7, 8).2.3 Docking simulationsStarting with RSCB deposited crystal structure of EGFR proteinhaving Erlotinib as a co-crystallized ligand (PDB: 1M17) (Bhat et al.,2022), and re-docking of Erlotinib revealed a docking score (S)of −7.30 kcal/mol and an RMSD of 1.28 Å, in addition to formationof the two characteristic H-bond interactions with two of key aminoacid residues, Gln767 and Met769, indicating validity of dockingstudy parameters. While running docking simulations within EGFRactive site (PDB ID: 1M17) showed that most of the test derivatives(9a-o) showed moderate to strong binding interactions(S = −5.93 to −7.52; c.f. Erlotinib: 7.30 kcal/mol) as listed inTable 3. These interactions were variable between H-bond and/orH-pi, with key amino acid residues (Met 769, Lys 721, Gly 772, andLeu 694) lining the active site, as shown in Supplementary Figure S1(Supplementary Material).Remarkably, derivative 9b (R1 = Ph. and R2 = Cl) showed ahigher docking score (S = −6.51 kcal/mol) over its methoxycongener (S = −6.04 kcal/mol) and has the best docking scoreover other p-chloro derivatives (9e, 9h, and 9k). Visualinspection of the docking pose with the lowest RMSD value andhighest docking score of compound 9b, we observed a stabilizingH-bonding and H-pi binding interactions through N-phenyl andamidic carbonyl group of quinazoline ring with Lys721 amino acidresidue, Figure 9. Such interactions were not found in other p-chloroFIGURE 6Cell cycle analysis results for compound 9b.FIGURE 7Apoptosis induction results of compound 9b.Frontiers in Chemistry frontiersin.org08Mohamed et al. 10.3389/fchem.2024.1447618derivatives, 9e and 9k (except with derivative 9h), because of thehydrophilic tale of Gln738 amino acid residue, which repels closelyfound non-hydrophilic groups as methyl group of 9e and ethylgroup of 9k, as shown in Supplementary Figure S1.Whereas upon working on binding interactions within mutantEGFR (EGFRT790M; PDB ID: 2JIU) active site, the p-Cl derivative(9b) gave the highest docking score (S = −7.43 kcal/mol) among all15 derivatives tested, as shown in Table 3. Most test derivativescommonly interacted through H-bonds and/or H-pi interactionswith Lys 745 and Leu 718, as seen with docking poses of derivatives9b and 9h, Figure 10.Finally, docking scores of derivatives (9a-o) interactions withinBRAF (PDB ID: 5JRQ) (Umar et al., 2020) active site were high andso close to each other (S = −6.24 to −7.65 kcal/mol). Additionally,multiple interactions varying from the weak Pi-Pi interactions,through H-Pi to the strong H-bonds with either Phe 583, Val471, Asp 594, or Lys 483, as shown in Figure 11 of compound 9l.To summarize, all the 15 derivatives showed good bindingprofiles within target proteins EGFR, mutant EGFR (EGFRT790M),and BRAFV600E, as seen from their docking scores and interactionswith amino acid residues lining their active sites, and this could beused to explain the possible mechanism by which such class ofcompounds inhibit these proteins activity.2.4 Calculation of ADME propertiesThe drug-likeness of new compounds 9a-o was calculated usingthe SwissADME website (Daina et al., 2017) to predict theirtransport properties through membranes like GIT and/or BBB.All the test compounds obey Lipinski’s rule of five (RO5) withMLogP below 5, in addition to having both a topological polarsurface area below 140 Å2 and molar refractivity below 130,indicating their facile transport through cell membranes andhence better oral bioavailability (F), as shown in Table 4.3 ConclusionA novel set of quinazoline-4-one/1,2,4-oxadiazole hybrids (9a-o)was designed and synthesized as EGFR, EGFRT790M, and BRAFV600Einhibitors in the search for multitargeted antiproliferative scaffold.The novel hybrids showed encouraging antiproliferative actions.Compounds 9b, 9c, 9h, 9k, and 9l were evaluated as EGFR andBRAFV600E inhibitors. These in vitro experiments demonstrated thatcompounds 9b, 9c, and 9h are potent antiproliferative agents capableof acting as dual EGFR/BRAFV600E inhibitors. 9b, 9c, and 9h werefurther studied for their inhibitory effect on mutant EGFR(EGFRT790M), with the results indicating that the evaluatedcompounds had a significant inhibitory effect. Cell cycle analysisand apoptosis induction assay of 9b revealed cell cycle arrest at the G2/M phase, which can induce apoptosis. EGFR and BRAFV600E dockingsimulations inside their active regions shed light on these compounds’possible modes of inhibition. ADME calculations revealed that all testcompounds satisfy Lipinski’s rule of five (RO5) with MLogP <5, witheasy transport through cell membranes and higher oral bioavailability.These new hybrids may have potential as anti-cancer drugs afteroptimization.FIGURE 8Cell cycle analysis and apoptosis induction results of compound 9b.Frontiers in Chemistry frontiersin.org09Mohamed et al. 10.3389/fchem.2024.14476184 Experimental4.1 Chemistry4.1.1 General detailsThe starting materials, quinazolines 3a-e (Moussa et al., 2018)and 3-aryl-5-(chloromethyl)-1,2,4-oxadiazole derivatives,compounds 8a-c (Minin et al., 2023), were prepared according toliterature methods.4.1.2 General procedures for the synthesis ofcompounds (9a-o)To a stirred solution of quinazoline derivatives (0.60 mmol,1 eq), compounds 3a-e, in DMF (5 mL), anhydrous K2CO3(0.72 mmol, 1.20 eq, 0.10 g) was added and stirred for 1h atroom temperature. Then, 3-aryl-5-(chloromethyl)-1,2,4-oxadiazole derivatives, compounds 8a-c, (0.60 mmol, 1 eq.)was added, and KI (0.60 mmol, 1 eq, 0.10 g) was also addedto the reaction mixture and stirring was continued for 24 h. Aftercompletion of the reaction (checked by TLC using Hexane: Ethylacetate 2:1), the reaction mixture was poured into crushed icewhile stirring. The obtained precipitate was filtered off, washedseveral times with water, dried at 60°C, and crystallizedfrom ethanol.4.1.3 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-phenylquinazolin-4(3H)-one (9a)Yield: 0.21 g (85%), White solid, m.p: 162°C–164°C, Rf. 0.66(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.07 (d, J = 7.7 Hz, 1H, Ar-H), 7.99 (d, J = 6.4 Hz,2H, Ar-H), 7.81 (t, J = 7.4 Hz, 1H, Ar-H), 7.60 (d, J = 11.0 Hz,4H, Ar-H), 7.57–7.52 (m, 4H, Ar-H), 7.47 (t, J = 8.0 Hz, 2H, Ar-H), 4.79 (s, 2H, S-CH2);13C NMR (100 MHz, δ ppm DMSO-d6):177.09, 167.84, 160.56, 155.46, 146.85, 135.49, 135.01, 131.59,130.22, 129.67, 129.46, 129.25, 126.98, 126.61, 126.32, 126.03,125.90, 119.60, 26.97; Anal. Calc. (%) For C23H16N4O2S: C,66.97; H, 3.91; N, 13.58; S, 7.77. Found: C, 66.81; H, 3.85; N,13.82; S, 7.85.4.1.4 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3 phenylquinazolin-4(3H)-one (9b)Yield: 0.24 g (89%), White solid, m.p: 172°C–174°C, Rf. 0.67(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.07 (dd, J = 8.2, 1.5 Hz, 1H, Ar-H), 8.00 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.84–7.79 (m, 1H, Ar-H), 7.63 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.62–7.60 (m, 3H, Ar-H), 7.57–7.52 (m, 2H, Ar-H),7.51–7.44 (m, 2H, Ar-H), 4.78 (s, 2H, S-CH2);13C NMR (100MHz, δppm DMSO-d6): 177.84, 167.52, 161.02, 155.92, 147.31, 136.82,135.95, 135.47, 130.70, 130.14, 129.93, 129.89, 129.24, 127.08,TABLE 3 Oxadiazoles docking scores in EGFR, EGFRT790M, and BRAFV600E active sites.Comp R1 R2 Docking score (S; kcal/mol)EGFR (1M17) EGFRT790M (2JIU) BRAFV600E (5JRQ)9a Phenyl H −6.0118 −7.1900 −7.04589b Phenyl Cl −6.5073 −7.4268 −7.09859c Phenyl OMe −6.0388 −7.2876 −7.22789d p-Tolyl H −6.0314 −6.7971 −6.92329e p-Tolyl Cl −6.4310 −5.6713 −7.30789f p-Tolyl OMe −6.8251 −6.7228 −6.99719g m-Tolyl H −7.5281 −6.8538 −6.86469h m-Tolyl Cl −6.2685 −6.4894 −6.84709i m-Tolyl OMe −6.8622 −6.7373 −7.65169j Ethyl H −6.6348 −6.9894 −6.28299k Ethyl Cl −5.7431 −6.3041 −6.58249l Ethyl OMe −5.9343 −6.1808 −6.70299m Allyl H −6.1516 −7.2311 −6.83909n Allyl Cl −6.2599 −6.3498 −6.84759o Allyl OMe −6.7193 −6.7397 −6.8732Erlotinib −7.3034 — —Osimertinib — −8.5638 —Vemurafenib — --- −9.3319Frontiers in Chemistry frontiersin.org10Mohamed et al. 10.3389/fchem.2024.1447618126.79, 126.36, 125.33, 120.08, 27.44; Anal. Calc. (%) ForC23H15ClN4O2S: C, 61.81; H, 3.38; N, 12.54; S, 7.17. Found: C,61.97; H, 3.50; N, 12.71; S, 7.28.4.1.5 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-phenylquinazolin-4(3H)-one (9c)Yield: 0.23 g (88%), White solid, m.p: 186°C–188°C, Rf.0.65 Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.07 (d, J = 7.7 Hz, 1H, Ar-H), 7.92 (d, J = 8.7 Hz, 2H,Ar-H p-OCH3 C6H4), 7.81 (t, J = 7.3 Hz, 1H, Ar-H), 7.62 (d, J =3.9 Hz, 3H, Ar-H), 7.59–7.35 (m, 4H, Ar-H), 7.09 (d, J = 8.7 Hz, 2H,Ar-H p-OCH3 C6H4), 4.76 (s, 2H, S-CH2), 3.82 (s, 3H, OCH3);13CNMR (100 MHz, δ ppm DMSO-d6): 176.64, 167.52, 161.71, 160.54,155.46, 146.85, 135.48, 134.99, 130.20, 129.64, 129.45, 128.67,126.60, 126.30, 125.89, 119.60, 118.28, 114.63, 55.36, 26.89; Anal.Calc. (%) For C24H18N4O3S: C, 65.14; H, 4.10; N, 12.66; S, 7.25.Found: C, 64.91; H, 4.27; N, 12.89; S, 7.23.4.1.6 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-p-tolylquinazolin-4(3H)-one (9d)Yield: 0.22 g (86%), White solid, m.p: 168°C–170°C, Rf. 0.69(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.06 (d, J = 7.5 Hz, 1H, Ar-H), 7.98 (d, J = 6.5 Hz, 2H, Ar-H),7.81 (t, J = 7.4 Hz, 3H, Ar-H), 7.59–7.52 (m, 4H, Ar-H), 7.47 (d, J =6.1 Hz, 1H, Ar-H), 7.39 (q, J = 8.2 Hz, 2H, Ar-H), 4.76 (s, 2H,S-CH2), 2.42 (s, 3H, CH3);13C NMR (100 MHz, δ ppm DMSO-d6):177.12, 167.83, 160.62, 155.78, 146.87, 139.99, 135.01, 132.86,131.63, 130.16, 129.29, 129.18, 126.99, 126.63, 126.32, 126.03,125.89, 119.60, 26.96, 20.89; Anal. Calc. (%) For C24H18N4O2S:C, 67.59; H, 4.25; N, 13.14; S, 7.52. Found: C, 67.34; H, 4.43; N,13.40; S, 7.68.4.1.7 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-p-tolylquinazolin-4(3H)-one (9e)Yield: 0.25 g (90%), White solid, m.p: 156°C–158°C, Rf. 0.70(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.05 (d, J = 6.9 Hz, 1H, Ar-H), 7.98 (d, J = 7.5 Hz, 2H, Ar-H p-ClC6H4), 7.80 (t, J = 6.4 Hz, 1H, Ar-H), 7.62 (d, J = 7.5 Hz, 2H, Ar-Hp-Cl C6H4), 7.45 (d, J = 7.5 Hz, 2H, Ar-H), 7.42–7.31 (m, 4H, Ar-H),4.76 (s, 2H, S-CH2), 2.42 (s, 3H, CH3);13C NMR (100 MHz, δ ppmDMSO-d6): 177.42, 167.06, 160.62, 155.76, 146.86, 140.00, 136.36,135.02, 132.84, 130.17, 129.47, 129.17, 128.80, 126.62, 126.34,125.89, 124.87, 119.59, 26.99, 20.90; Anal. Calc. (%) ForC24H17ClN4O2S: C, 62.54; H, 3.72; N, 12.16; S, 6.96. Found: C,62.37; H, 3.80; N, 12.42; S, 6.89.4.1.8 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-p-tolylquinazolin-4(3H)-one (9f)Yield: 0.24 g (89%), White solid, m.p: 165°C–167°C, Rf. 0.68(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.06 (d, J = 7.6 Hz, 1H, Ar-H), 7.91 (d, J = 8.5 Hz, 2H,Ar-H p-OCH3 C6H4), 7.80 (t, J = 7.4 Hz, 1H, Ar-H), 7.48 (d, J =7.8 Hz, 2H, Ar-H), 7.39 (q, J = 7.7 Hz, 4H, Ar-H), 7.09 (d, J =8.5 Hz, 2H, Ar-H p-OCH3 C6H4), 4.74 (s, 2H, S-CH2), 3.82 (s, 3H,OCH3), 2.42 (s, 3H, CH3);13C NMR (100 MHz, δ ppm DMSO-d6): 176.65, 167.54, 161.73, 160.62, 155.74, 146.87, 139.98, 134.97,FIGURE 9Binding Interactions of 9b within EGFR (PDB ID: 1M17) active site showing both H-bonds and H-Pi interactions as green-dotted arrows and lines,respectively.Frontiers in Chemistry frontiersin.org11Mohamed et al. 10.3389/fchem.2024.1447618132.86, 130.15, 129.17, 128.68, 126.62, 126.29, 125.90, 119.59,118.30, 114.65, 55.38, 26.90, 20.90; Anal. Calc. (%) ForC25H20N4O3S: C, 65.77; H, 4.42; N, 12.27; S, 7.02. Found: C,65.62; H, 4.61; N, 12.41; S, 6.98.4.1.9 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-m-tolylquinazolin-4(3H)-one (9g)Yield: 0.22 g (86%), White solid, m.p: 170°C–172°C, Rf. 0.69(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-FIGURE 10Docking poses of compound 9b (top figure) and 9h (bottom figure) within the active site of EGFRT790M (PDB ID: 2JIU) showing H-Pi interactions withkey amino acid residues.Frontiers in Chemistry frontiersin.org12Mohamed et al. 10.3389/fchem.2024.1447618d6): 8.06 (dd, J = 8.2, 1.4 Hz, 1H, Ar-H), 7.98 (dd, J = 7.9, 1.7 Hz, 2H,Ar-H), 7.82–7.78 (m, 1H, Ar-H), 7.59–7.54 (m, 3H, Ar-H),7.49–7.45 (m, 3H, Ar-H), 7.41 (d, J = 7.7 Hz, 1H, Ar-H), 7.31 (d,J = 8.7 Hz, 2H, Ar-H), 4.76 (s, 2H, S-CH2), 2.40 (s, 3H, CH3);13CNMR (100 MHz, δ ppm DMSO-d6): 177.15, 167.86, 160.57, 155.56,146.87, 139.34, 135.40, 135.07, 131.65, 130.91, 129.65, 129.49,129.31, 127.01, 126.63, 126.44, 126.38, 126.04, 125.92, 119.60,26.99, 20.79; Anal. Calc. (%) For C24H18N4O2S: C, 67.59; H, 4.25;N, 13.14; S, 7.52. Found: C, 67.36; H, 4.09; N, 13.41; S, 7.60.4.1.10 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-m-tolylquinazolin-4(3H)-one (9h)Yield: 0.25 g (90%), White solid, m.p: 178°C–180°C, Rf. 0.70(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.09 (d, J = 7.7 Hz, 1H, Ar-H), 8.02 (d, J = 8.6 Hz, 2H,Ar-H p-Cl C6H4), 7.84 (t, J = 8.4 Hz, 1H, Ar-H), 7.66 (d, J =8.6 Hz, 2H, Ar-H p-Cl C6H4), 7.60–7.41 (m, 4H, Ar-H), 7.34 (d,J = 10.4 Hz, 2H, Ar-H), 4.79 (s, 2H, S-CH2), 2.43 (s, 3H, CH3);13CNMR (100 MHz, δ ppm DMSO-d6): 177.37, 167.03, 160.50,155.49, 146.81, 139.28, 136.33, 135.35, 134.99, 130.86, 129.61,129.43, 129.23, 128.76, 126.58, 126.40, 126.31, 125.86, 124.85,119.57, 26.95, 20.75; Anal. Calc. (%) For C24H17ClN4O2S: C, 62.54; H, 3.72; N, 12.16; S, 6.96. Found: C, 62.39; H, 3.85; N,12.40; S, 6.89.4.1.11 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-m-tolylquinazolin-4(3H)-one (9i)Yield: 0.24 g (89%), White solid, m.p: 194°C–196°C, Rf .0.68(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.09 (d, J = 7.6 Hz, 1H, Ar-H), 7.95 (d, J = 8.7 Hz, 2H, Ar-Hp-OCH3 C6H4), 7.84 (t, J = 7.3 Hz, 1H, Ar-H), 7.59–7.47 (m, 3H, Ar-H), 7.44 (d, J = 7.6 Hz, 1H, Ar-H), 7.34 (d, J = 10.8 Hz, 2H, Ar-H),FIGURE 112D-Binding interactions of compound 9l within BRAFV600E (PDB ID: 5JRQ) active site showing H-bonds (as blue-arrows) and pi-pi interactions (asgreen-dotted lines) with Asp 594 and Phe 583, respectively.Frontiers in Chemistry frontiersin.org13Mohamed et al. 10.3389/fchem.2024.14476187.12 (d, J = 8.7 Hz, 2H, Ar-H p-OCH3 C6H4), 4.77 (s, 2H, S-CH2),3.85 (s, 3H, OCH3), 2.43 (s, 3H, CH3);13C NMR (100 MHz, δ ppmDMSO-d6): 176.66, 167.51, 161.72, 160.52, 155.52, 146.83, 139.28,135.37, 135.00, 130.85, 129.62, 129.43, 128.67, 126.58, 126.41,126.31, 125.89, 119.57, 118.28, 114.65, 55.37, 26.90, 20.75; Anal.Calc. (%) For C25H20N4O3S: C, 65.77; H, 4.42; N, 12.27; S, 7.02.Found: C, 65.62; H, 4.61; N, 12.49; S, 7.14.4.1.12 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-ethylquinazolin-4(3H)-one (9j)Yield: 0.17 g (79%), White solid, m.p: 114°C–116°C, Rf. 0.60(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.04 (d, J = 8.0 Hz, 1H, Ar-H), 7.97 (d, J = 6.4 Hz, 2H,Ar-H), 7.74 (t, J = 7.4 Hz, 1H, Ar-H), 7.57–7.54 (m, 3H, Ar-H), 7.42 (t,J = 7.4 Hz, 1H, Ar-H), 7.35 (d, J = 8.4 Hz, 1H, Ar-H), 4.94 (s, 2H,S-CH2), 4.11 (q, J = 7.0 Hz, 2H, N-CH2), 1.31 (t, J = 7.0 Hz, 3H, CH3);13C NMR (100MHz, δ ppmDMSO-d6): 177.21, 167.84, 160.13, 154.45,146.41, 134.73, 131.60, 129.27, 126.95, 126.38, 126.22, 126.00, 125.66,118.84, 26.70, 13.04; Anal. Calc. (%) For C19H16N4O2S: C, 62.62; H,4.43; N, 15.37; S, 8.80. Found: C, 62.89; H, 4.51; N, 15.62; S, 8.71.4.1.13 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-ethylquinazolin-4(3H)-one (9k)Yield: 0.20 g (84%), White solid, m.p: 118°C–120°C, Rf. 0.62(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.04 (dd, J = 7.9, 1.0 Hz, 1H, Ar-H), 7.98 (d, J = 8.6 Hz, 2H, Ar-Hp-Cl C6H4), 7.75–7.72 (m, 1H, Ar-H), 7.60 (d, J = 8.6 Hz, 2H, Ar-Hp-Cl C6H4), 7.44–7.41 (m, 1H, Ar-H), 7.35 (d, J = 8.1 Hz, 1H, Ar-H),4.94 (s, 2H, S-CH2), 4.11 (q, J = 7.0 Hz, 2H, N-CH2), 1.31 (t, J =7.0 Hz, 3H, CH3);13C NMR (100 MHz, δ ppm DMSO-d6): 177.49,167.05, 160.11, 154.42, 146.39, 136.35, 134.72, 129.43, 128.73,126.37, 126.22, 125.64, 124.82, 118.83, 39.58, 26.67, 13.03; Anal.Calc. (%) For C19H15ClN4O2S: C, 57.21; H, 3.79; N, 14.05; S, 8.04.Found: C, 57.49; H, 3.86; N, 14. 27; S, 8.12.4.1.14 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-ethylquinazolin-4(3H)-one (9l)Yield: 0.20 g (84%), White solid, m.p: 134°C–136°C, Rf. 0.57(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppmDMSO-d6): 8.04 (dd, J = 8.0, 1.3 Hz, 1H, Ar-H), 7.91 (d, J = 8.9 Hz,2H, Ar-H p-OCH3 C6H4), 7.76–7.71 (m, 1H, Ar-H), 7.44–7.40 (m, 1H,Ar-H), 7.37 (d, J = 7.6 Hz, 1H, Ar-H), 7.07 (d, J = 8.9 Hz, 2H, Ar-Hp-OCH3 C6H4), 4.91 (s, 2H, S-CH2), 4.11 (q, J = 7.1 Hz, 2H, N-CH2),3.81 (s, 3H, OCH3), 1.31 (t, J = 7.1 Hz, 3H, CH3);13C NMR (100MHz, δppm DMSO-d6): 176.74, 167.52, 161.71, 160.11, 154.42, 146.40, 134.69,128.63, 126.35, 126.9, 125.65, 118.83, 118.25, 114.63, 55.35, 39.56, 26.62,13.01; Anal. Calc. (%) For C20H18N4O3S: C, 60.90; H, 4.60; N, 14.20; S,8.13. Found: C, 61.13; H, 4.74; N, 14.37; S, 8.20.4.1.15 2-((3-Phenyl-1,2,4-oxadiazol-5-yl)methylthio)-3-allylquinazolin-4(3H)-one (9m)Yield: 0.18 g (80%),White solid, m.p: 110°C–112°C, Rf. 0.61(Hexane:Ethyl acetate, 2:1, v/v). 1H NMR (400 MHz, δ ppm DMSO-d6): 8.04 (d,J = 7.8 Hz, 1H, Ar-H), 7.97 (d, J = 6.8 Hz, 2H, Ar-H), 7.75 (t, J = 7.5 Hz,1H, Ar-H), 7.62–7.49 (m, 3H, Ar-H), 7.43 (t, J = 7.5 Hz, 1H, Ar-H), 7.37(d, J = 8.1 Hz, 1H, Ar-H), 6.11–5.83 (m, 1H, =CH), 5.24 (d, J = 10.4 Hz,TABLE 4 Theoretical calculations of ADME properties of compounds (9a-o) using swissADME software.Cpd MW Lipinski Parametersa F MR Water SolubilitySilicos-IT classGI absorption BBBHBA HBD nrotb TPSA MLogP9a 412.46 5 0 5 99.11 3.6 0.55 116.84 Poor High No9b 446.91 5 0 5 99.11 4.08 0.55 121.85 Poor High No9c 442.49 6 0 6 108.34 3.28 0.55 123.33 Poor High No9d 426.49 5 0 5 99.11 3.81 0.55 121.81 Poor High No9e 460.94 5 0 5 99.11 4.29 0.55 126.82 Poor High No9f 456.52 6 0 6 108.34 3.49 0.55 128.3 Poor High No9g 426.49 5 0 5 99.11 3.81 0.55 121.81 Poor High No9h 460.94 5 0 5 99.11 4.29 0.55 126.82 Poor High No9i 456.52 6 0 6 108.34 3.49 0.55 128.3 Poor High No9j 364.42 5 0 5 99.11 2.92 0.55 101.57 Poor High No9k 398.87 5 0 5 99.11 3.15 0.55 106.58 Poor High No9l 394.45 6 0 6 108.34 2.35 0.55 108.06 Poor High No9m 376.43 5 0 6 99.11 3.08 0.55 105.9 Poor High No9n 410.88 5 0 6 99.11 3.3 0.55 110.91 Poor High No9o 406.46 6 0 7 108.34 2.5 0.55 112.4 Poor High NoMW, molecular weight; HBA, H-bond acceptor; HBD, H-bond donor; nrotb, no. of rotatable bonds; TPSA, topological polar surface area (Å2); MLogP, n-octanol/water distribution coefficient;F, Abbott bioavailability scores (0–1).aDrug lead-like character: MW ≤ 500, HBA ≤10, HBD ≤5, nrotb ≤10, TPSA ≤140, lipophilicity parameter MLogP ≤5; MR, 40–130; BBB, Blood-Brain Barrier.Frontiers in Chemistry frontiersin.org14Mohamed et al. 10.3389/fchem.2024.14476181H, =CH2), 5.15 (d, J = 17.3 Hz, 1H, =CH2), 4.92 (s, 2H, S-CH2), 4.72 (d,J = 4.0 Hz, 2H, N-CH2);13C NMR (100MHz, δ ppmDMSO-d6): 177.20,167.84, 160.19, 154.93, 146.44, 134.94, 131.64, 131.25, 129.30, 126.96,126.53, 126.37, 126.01, 125.74, 118.73, 117.60, 45.97, 26.80; Anal. Calc.(%) For C20H16N4O2S: C, 63.81; H, 4.28; N, 14.88; S, 8.52. Found: C,64.05; H, 4.42; N, 15.17; S, 8.64.4.1.16 2-((3-(4-Chlorophenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-allylquinazolin-4(3H)-one (9n)Yield: 0.21 g (85%), White solid, m.p: 115°C–117°C, Rf. 0.63(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.07 (dd, J = 8.0, 1.2 Hz, 1H, Ar-H), 7.99 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.80–7.74 (m, 1H, Ar-H), 7.63 (d, J = 8.7 Hz, 2H, Ar-Hp-Cl C6H4), 7.49–7.43 (m, 1H, Ar-H), 7.39 (d, J = 8.2, 1H, Ar-H),6.04–5.90 (m, 1H, =CH), 5.26 (dd, J = 10.4, 1.2 Hz, 1H, =CH2), 5.17(dd, J = 17.2, 1.2 Hz, 1H, =CH2), 4.93 (s, 2H, S-CH2), 4.74 (d, J =5.1 Hz, 2H, N-CH2);13C NMR (100MHz, δ ppmDMSO-d6): 177.39,167.02, 160.10, 154.80, 146.37, 136.32, 134.81, 131.19, 129.38,128.68, 126.47, 126.27, 125.68, 124.80, 118.69, 117.61, 45.92,26.72; Anal. Calc. (%) For C20H15ClN4O2S: C, 58.46; H, 3.68; N,13.64; S, 7.80. Found: C, 58.70; H, 3.73; N, 13. 91; S, 7. 94.4.1.17 2-((3-(4-Methoxyphenyl)-1,2,4-oxadiazol-5-yl)methylthio)-3-allylquinazolin 4(3H)-one (9o)Yield: 0.20 g (83%), White solid, m.p: 122°C–124°C, Rf. 0.59(Hexane: Ethyl acetate, 2:1, v/v). 1H NMR (400MHz, δ ppmDMSO-d6): 8.07 (d, J = 7.6 Hz, 1H, Ar-H), 7.92 (d, J = 8.7 Hz, 2H, Ar-Hp-OCH3 C6H4), 7.78 (t, J = 7.2 Hz, 1H, Ar-H), 7.46 (t, J = 7.4 Hz, 1H,Ar-H), 7.40 (d, J = 8.4 Hz, 1H, Ar-H), 7.09 (d, J = 8.7 Hz, 2H, Ar-Hp-OCH3 C6H4), 6.01–5.91 (m, 1H, =CH), 5.26 (d, J = 10.4 Hz,1H, =CH2), 5.17 (d, J = 17.3 Hz, 1H, =CH2), 4.91 (s, 2H, S-CH2), 4.74(d, J = 4.7 Hz, 2H, 2H, N-CH2), 3.82 (s, 3H, OCH3);13C NMR(100MHz, δ ppmDMSO-d6): 176.73, 167.53, 161.73, 160.16, 154.89,146.43, 134.88, 131.24, 128.64, 126.51, 126.33, 125.74, 118.72,118.28, 117.58, 114.66, 55.37, 45.93, 26.70; Anal. Calc. (%) ForC21H18N4O3S: C, 62.05; H, 4.46; N, 13.78; S, 7.89. Found: C,62.29; H, 4.53; N, 14.02; S, 7.81.4.2 Biology4.2.1 Assay of cell viability of 9a-oThe human mammary gland epithelial (MCF-10A) normal cellline was used to test the viability of compounds 9a-o (Mahmoudet al., 2022; Mekheimer et al., 2022). Refer to SupplementaryAppendix A (Supplementary Material) for more details.4.2.2 Assay of antiproliferative actionThe MTT assay was used to investigate the antiproliferativeactivity of hybrids 9a-o versus four human cancer cell lines usingErlotinib as a control: colon cancer (HT-29) cell line, pancreaticcancer (Panc-1) cell line, lung cancer (A-549) cell line, and breastcancer (MCF-7) cell line (El-Sherief et al., 2019; Al-Wahaibi et al.,2022). See Supplementary Appendix A for more information.4.2.3 EGFR inhibitory assayThe EGFR-TK test (Abdel-Aziz et al., 2023) assessed theinhibitory potency of the most effective antiproliferativederivatives 9b, 9c, 9h, 9k, and 9l against EGFR. For more details,see Supplementary Appendix A.4.2.4 BRAFV600E inhibitory assayAn in vitro investigation assessed the anti-BRAFV600E activity of9b, 9c, 9h, 9k, and 9l (Youssif et al., 2022). Refer to SupplementaryAppendix A for more details.4.2.5 EGFRT790M inhibitory assayThe HTRF KinEASE-TK assay (Miles et al., 2020) was used toevaluate the inhibitory action of the most potent hybrids, 9b, 9c, and9h, against mutant-type EGFR (EGFRT790M). For more details, seeSupplementary Appendix A.4.2.6 Cell cycle analysis and apoptosis detectionCompound 9b was investigated for its effects on cell cycleprogression and apoptosis induction in A-549 cells. A lungcancer (A-549) cell line was treated for 24 h with an IC50concentration of 9b. The cell line was labeled with PI/AnnexinV, and flow cytometry was done with a BD FASC Caliber (El-Sherief et al., 2018). See Supplementary Appendix A formore details.4.3 Docking studyMolecular docking simulations of 15 derivatives (9a-o) wereperformed via Molecular Operating Environment (MOE®)software according to reported protocols (Abdel-Aziz et al.,2023) within the active site of EGFR tyrosine kinase domain(PDB ID: 1M17), mutant EGFR kinase domain T790M(EGFRT790M; PDB ID: 2JIU), and mutant BRAF kinase domain(BRAFV600E; PDB ID: 5JRQ) crystals structures downloaded fromRSCB protein data bank (https://www.rcsb.org/). For more details,see Supplementary Appendix A.4.4 Calculations of SwissADMEPharmacokinetics and drug-likeness prediction for all the newlysynthesized compounds was performed using the online toolSwissADME predictor software (http://www.swissadme.ch/) madeby the Swiss Institute of Bioinformatics.Data availability statementData will be available upon request from the authors.Author contributionsAM: Formal Analysis, Methodology, Software, Writing–originaldraft. OA-G: Investigation, Supervision, Validation, Visualization,Writing–original draft. YM: Methodology, Supervision,Writing–original draft. KD: Funding acquisition, Writing–originaldraft. SB: Investigation, Validation, Visualization, Writing–reviewand editing. BY: Conceptualization, Data curation, Formal Analysis,Frontiers in Chemistry frontiersin.org15Mohamed et al. 10.3389/fchem.2024.1447618Investigation, Methodology, Resources, Software, Supervision,Validation, Visualization, Writing–original draft, Writing–reviewand editing.FundingThe author(s) declare that financial support was received for theresearch, authorship, and/or publication of this article. This workwas funded by the Researchers Supporting Project Number(RSP2024R388) King Saud University, Riyadh, Saudi Arabia. Theauthors also acknowledge support from the KIT-Publication Fundof the Karlsruhe Institute of Technology.AcknowledgmentsThis work was funded by the Researchers Supporting ProjectNumber (RSP2024R388) King Saud University, Riyadh, SaudiArabia. The authors also acknowledge support from the KIT-Publication Fund of the Karlsruhe Institute of Technology.Conflict of interestThe authors declare that the research was conducted in theabsence of any commercial or financial relationships that could beconstrued as a potential conflict of interest.Publisher’s noteAll claims expressed in this article are solely those of the authors anddo not necessarily represent those of their affiliated organizations, orthose of the publisher, the editors and the reviewers. Any product thatmay be evaluated in this article, or claim that may be made by itsmanufacturer, is not guaranteed or endorsed by the publisher.Supplementary materialThe Supplementary Material for this article can be found onlineat: https://www.frontiersin.org/articles/10.3389/fchem.2024.1447618/full#supplementary-materialReferencesAbdel-Aziz, M., Beshr, E. A., Gomaa, H. A., Mostafa, Y. A., Youssif, B. G., Hayallah, A.M., et al. (2023). Discovery of new cyanopyridine/chalcone hybrids as dual inhibitors ofEGFR/BRAFV600E with promising antiproliferative properties. Arch. Pharm. 356 (4),2200464. doi:10.1002/ardp.202200464Alshammari, M. B., Aly, A. A., Youssif, B. G., Bräse, S., Ahmad, A., Brown, A. B., et al.(2022). Design and synthesis of new thiazolidinone/uracil derivatives asantiproliferative agents targeting EGFR and/or BRAFV600E. Front. Chem. 10,1076383. doi:10.3389/fchem.2022.1076383Al-Wahaibi, L. H., Abou-Zied, H. A., Beshr, E. A., Youssif, B. G., Hayallah, A. M., andAbdel-Aziz, M. (2023a). Design, synthesis, antiproliferative actions, and DFT studies ofnew bis–pyrazoline derivatives as dual EGFR/BRAFV600E inhibitors. Int. J. Mol. Sci. 24(10), 9104. doi:10.3390/ijms24109104Al-Wahaibi, L. H., Abou-Zied, H. A., Hisham, M., Beshr, E. A., Youssif, B. G., Bräse,S., et al. (2023b). Design, synthesis, and biological evaluation of novel 3-cyanopyridone/pyrazoline hybrids as potential apoptotic antiproliferative agents targeting EGFR/BRAFV600E inhibitory pathways. Molecules 28 (18), 6586. doi:10.3390/molecules28186586Al-Wahaibi, L. H., El-Sheref, E. M., Hammouda, M. M., and Youssif, B. G. (2023c).One-pot synthesis of 1-Thia-4-azaspiro [4.4/5] alkan-3-ones via schiff base: design,synthesis, and apoptotic antiproliferative properties of dual EGFR/BRAFV600Einhibitors. Pharmaceuticals 16 (3), 467. doi:10.3390/ph16030467Al-Wahaibi, L. H., El-Sheref, E. M., Hassan, A. A., Bräse, S., Nieger, M., Youssif, B. G.,et al. (2023d). Synthesis and structure determination of substituted thiazole derivativesas EGFR/BRAFV600E dual inhibitors endowed with antiproliferative activity.Pharmaceuticals 16 (7), 1014. doi:10.3390/ph16071014Al-Wahaibi, L. H., Gouda, A.M., Abou-Ghadir, O. F., Salem, O. I., Ali, A. T., Farghaly,H. S., et al. (2020). Design and synthesis of novel 2, 3-dihydropyrazino [1, 2-a] indole-1,4-dione derivatives as antiproliferative EGFR and BRAFV600E dual inhibitors. Bioorg.Chem. 104, 104260. doi:10.1016/j.bioorg.2020.104260Al-Wahaibi, L. H., Mahmoud, M. A., Mostafa, Y. A., Raslan, A. E., and Youssif, B. G.(2023e). Novel piperine-carboximidamide hybrids: design, synthesis, andantiproliferative activity via a multi-targeted inhibitory pathway. J. EnzymeInhibition Med. Chem. 38 (1), 376–386. doi:10.1080/14756366.2022.2151593Al-Wahaibi, L. H., Mostafa, Y. A., Abdelrahman, M. H., El-Bahrawy, A. H.,Trembleau, L., and Youssif, B. G. (2022). Synthesis and biological evaluation ofindole-2-carboxamides with potent apoptotic antiproliferative activity as EGFR/CDK2 dual inhibitors. Pharmaceuticals 15 (8), 1006. doi:10.3390/ph15081006Benassi, A., Doria, F., and Pirota, V. (2020). Groundbreaking anticancer activity of highlydiversified oxadiazole scaffolds. Int. J. Mol. Sci. 21 (22), 8692. doi:10.3390/ijms21228692Bhat, M. A., Tüzün, B., Alsaif, N. A., Khan, A. A., and Naglah, A. M. (2022). Synthesis,characterization, molecular modeling against EGFR target and ADME/T analysis ofnovel purine derivatives of sulfonamides. J. Mol. Struct. 1257, 132600. doi:10.1016/j.molstruc.2022.132600Boran, A. D., and Iyengar, R. (2010). Systems approaches to polypharmacology anddrug discovery. Curr. Opin. drug Discov. & Dev. 13 (3), 297–309.Daina, A., Michielin, O., and Zoete, V. (2017). SwissADME: a free web tool to evaluatepharmacokinetics, drug-likeness and medicinal chemistry friendliness of smallmolecules. Sci. Rep. 7 (1), 42717. doi:10.1038/srep42717Derbyshire, E. R., Mazitschek, R., and Clardy, J. (2012). Characterization ofPlasmodium liver stage inhibition by halofuginone. ChemMedChem 7 (5), 844–849.doi:10.1002/cmdc.201200045Do, B., Mace, M., and Rexwinkle, A. (2016). Idelalisib for treatment of B-cellmalignancies. Am. J. health-system Pharm. 73 (8), 547–555. doi:10.2146/ajhp150281El Mansouri, A.-E., Oubella, A., Maatallah, M., AitItto, M. Y., Zahouily, M., Morjani,H., et al. (2020). Design, synthesis, biological evaluation and molecular docking of newuracil analogs-1, 2, 4-oxadiazole hybrids as potential anticancer agents. Bioorg. & Med.Chem. Lett. 30 (19), 127438. doi:10.1016/j.bmcl.2020.127438El-Sherief, H. A., Youssif, B. G., Abdelazeem, A. H., Abdel-Aziz, M., and Abdel-Rahman, H. M. (2019). Design, synthesis and antiproliferative evaluation of novel 1, 2,4-triazole/schiff base hybrids with EGFR and B-RAF inhibitory activities. Anti-CancerAgents Med. Chem. Former. Curr. Med. Chemistry-Anti-Cancer Agents. 19 (5), 697–706.doi:10.2174/1871520619666181224115346El-Sherief, H. A., Youssif, B. G., Bukhari, S. N. A., Abdelazeem, A. H., Abdel-Aziz, M.,and Abdel-Rahman, H. M. (2018). Synthesis, anticancer activity and molecularmodeling studies of 1, 2, 4-triazole derivatives as EGFR inhibitors. Eur. J. Med.Chem. 156, 774–789. doi:10.1016/j.ejmech.2018.07.024Finn, R. S., Martin, M., Rugo, H. S., Jones, S., Im, S.-A., Gelmon, K., et al. (2016).Palbociclib and letrozole in advanced breast cancer. N. Engl. J. Med. 375 (20),1925–1936. doi:10.1056/nejmoa1607303Flaherty, K. T., Infante, J. R., Daud, A., Gonzalez, R., Kefford, R. F., Sosman, J., et al.(2012). Combined BRAF and MEK inhibition in melanoma with BRAFV600 mutations. N. Engl. J. Med. 367 (18), 1694–1703. doi:10.1056/nejmoa1210093Food, U., and Administration, D. (1997). Center for drug evaluation andresearch, guidance for industry: dissolution testing of immediate release solidoral dosage forms.Fu, R.-g., Sun, Y., Sheng, W.-b., and Liao, D.-f. (2017). Designing multi-targetedagents: an emerging anticancer drug discovery paradigm. Eur. J. Med. Chem. 136,195–211. doi:10.1016/j.ejmech.2017.05.016Hendawy, O. M. (2022). A comprehensive review of recent advances in the biologicalactivities of 1, 2, 4-oxadiazoles. Arch. Pharm. 355 (7), 2200045. doi:10.1002/ardp.202200045Hisham, M., Hassan, H. A., Gomaa, H. A., Youssif, B. G., Hayalah, A. M., and Abdel-Aziz, M. (2023). Design, synthesis, and antiproliferative activity of quinazolin-4-one/chalcone hybrids via the egfr inhibition pathway. Anti-Cancer Agents Med. Chem.Former. Curr. Med. Chemistry-Anti-Cancer Agents. 23 (17), 1932–1943. doi:10.2174/1871520623666230727104933Frontiers in Chemistry frontiersin.org16Mohamed et al. 10.3389/fchem.2024.1447618Hisham, M., Hassan, H. A., Gomaa, H. A., Youssif, B. G., Hayallah, A. M., and Abdel-Aziz, M. (2022). Structure-based design, synthesis and antiproliferative action of newquinazoline-4-one/chalcone hybrids as EGFR inhibitors. J. Mol. Struct. 1254, 132422.doi:10.1016/j.molstruc.2022.132422Keith, C. T., Borisy, A. A., and Stockwell, B. R. (2005). Multicomponent therapeuticsfor networked systems. Nat. Rev. Drug Discov. 4 (1), 71–78. doi:10.1038/nrd1609Kerru, N., Gummidi, L., Maddila, S., Gangu, K. K., and Jonnalagadda, S. B. (2020). Areview on recent advances in nitrogen-containing molecules and their biologicalapplications. Molecules 25 (8), 1909. doi:10.3390/molecules25081909Kitadai, N., and Maruyama, S. (2018). Origins of building blocks of life: a review.Geosci. Front. 9 (4), 1117–1153. doi:10.1016/j.gsf.2017.07.007Loboda, K. B., Valjavec, K., Štampar, M., Wolber, G., Žegura, B., Filipič, M., et al.(2020). Design and synthesis of 3, 5-substituted 1, 2, 4-oxadiazoles as catalytic inhibitorsof human DNA topoisomerase IIα. Bioorg. Chem. 99, 103828. doi:10.1016/j.bioorg.2020.103828Mahmoud, M. A., Mohammed, A. F., Salem, O. I., Gomaa, H. A., and Youssif, B. G.(2022). New 1, 3, 4-oxadiazoles linked with the 1, 2, 3-triazole moiety as antiproliferativeagents targeting the EGFR tyrosine kinase. Arch. Pharm. 355 (6), 2200009. doi:10.1002/ardp.202200009McLaughlin, N. P., Evans, P., and Pines, M. (2014). The chemistry and biology offebrifugine and halofuginone. Bioorg. & Med. Chem. 22 (7), 1993–2004. doi:10.1016/j.bmc.2014.02.040Meanwell, N. (2017). A synopsis of the properties and applications of heteroaromaticrings in medicinal chemistry. Adv. Heterocycl. Chem., 245–361. doi:10.1016/bs.aihch.2016.11.002Medina-Franco, J. L., Giulianotti, M. A.,Welmaker, G. S., andHoughten, R. A. (2013).Shifting from the single to the multitarget paradigm in drug discovery. Drug Discov.today 18 (9-10), 495–501. doi:10.1016/j.drudis.2013.01.008Mekheimer, R. A., Allam, S. M., Al-Sheikh, M. A., Moustafa, M. S., Al-Mousawi, S. M.,Mostafa, Y. A., et al. (2022). Discovery of new pyrimido [5, 4-c] quinolines as potentialantiproliferative agents with multitarget actions: rapid synthesis, docking, and ADMEstudies. Bioorg. Chem. 121, 105693. doi:10.1016/j.bioorg.2022.105693Miles, D., Chen, Y., Paprcka, S. L., Foley, C. N., Grange, R., Leleti, M. R., et al. (2020).Discovery and characterization of potent and selective AXL receptor tyrosine kinaseinhibitors for cancer therapy. Eur. J. Cancer 138, S38–S39. doi:10.1016/s0959-8049(20)31175-8Minin, D., Popkov, S., Pesochinskaya, K., and Aleksanov, D. (2023). Synthesis of [4-(3-aryl-1, 2, 4-oxadiazol-5-ylmethylthio)-2-methylphenoxy] acetic acids as new potential PPARagonists. Russ. Chem. Bull. 72 (7), 1648–1653. doi:10.1007/s11172-023-3944-5Mondaca, S., Lacouture, M., Hersch, J., and Yaeger, R. (2018). Balancing RAF, MEK,and EGFR inhibitor doses to achieve clinical responses and modulate toxicity in BRAFV600E colorectal cancer, 2. JCO precision oncol. doi:10.1200/PO.18.00088Moussa, G., Alaaeddine, R., Alaeddine, L. M., Nassra, R., Belal, A. S., Ismail, A., et al.(2018). Novel click modifiable thioquinazolinones as anti-inflammatory agents: design,synthesis, biological evaluation and docking study. Eur. J. Med. Chem. 144, 635–650.doi:10.1016/j.ejmech.2017.12.065Notarangelo, T., Sisinni, L., Condelli, V., and Landriscina, M. (2017). Dual EGFR andBRAF blockade overcomes resistance to vemurafenib in BRAF mutated thyroidcarcinoma cells. Cancer cell Int. 17 (1), 86–89. doi:10.1186/s12935-017-0457-zOkaniwa, M., Hirose, M., Imada, T., Ohashi, T., Hayashi, Y., Miyazaki, T., et al.(2012). Design and synthesis of novel DFG-out RAF/vascular endothelial growth factorreceptor 2 (VEGFR2) inhibitors. 1. Exploration of [5, 6]-fused bicyclic scaffolds. J. Med.Chem. 55 (7), 3452–3478. doi:10.1021/jm300126xPadwa, A., and Bur, S. K. (2007). The domino way to heterocycles. Tetrahedron 63(25), 5341–5378. doi:10.1016/j.tet.2007.03.158Paraiso, K., Fedorenko, I., Cantini, L., Munko, A., Hall, M., Sondak, V., et al. (2010).Recovery of phospho-ERK activity allows melanoma cells to escape from BRAFinhibitor therapy. Br. J. cancer 102 (12), 1724–1730. doi:10.1038/sj.bjc.6605714Purcell, J. W., Davis, J., Reddy, M., Martin, S., Samayoa, K., Vo, H., et al. (2010).Activity of the kinesin spindle protein inhibitor ispinesib (SB-715992) in models ofbreast cancer. Clin. Cancer Res. 16 (2), 566–576. doi:10.1158/1078-0432.ccr-09-1498Raghavendra, N. M., Pingili, D., Kadasi, S., Mettu, A., and Prasad, S. (2018). Dual ormulti-targeting inhibitors: the next generation anticancer agents. Eur. J. Med. Chem.143, 1277–1300. doi:10.1016/j.ejmech.2017.10.021Robert, C., Karaszewska, B., Schachter, J., Rutkowski, P., Mackiewicz, A., Stroiakovski,D., et al. (2015). Improved overall survival in melanoma with combined dabrafenib andtrametinib. N. Engl. J. Med. 372 (1), 30–39. doi:10.1056/nejmoa1412690Sharma, V., Kamal, R., and Kumar, V. (2017). Heterocyclic analogues as kinaseinhibitors: a focus review. Curr. Top. Med. Chem. 17 (22), 2482–2494. doi:10.2174/1568026617666170307113744Umar, A. B., Uzairu, A., Shallangwa, G. A., and Uba, S. (2020). QSAR modelling andmolecular docking studies for anti-cancer compounds against melanoma cell line SK-MEL-2. Heliyon 6 (3), e03640. doi:10.1016/j.heliyon.2020.e03640Unadkat, V., Rohit, S., Parikh, P., Sanna, V., and Singh, S. (2021). Rational design-aided discovery of novel 1, 2, 4-oxadiazole derivatives as potential EGFR inhibitors.Bioorg. Chem. 114, 105124. doi:10.1016/j.bioorg.2021.105124Wahid, M., Jawed, A., Mandal, R. K., Dar, S. A., Akhter, N., Somvanshi, P., et al.(2018). Recent developments and obstacles in the treatment of melanoma with BRAFand MEK inhibitors. Crit. Rev. oncology/hematology 125, 84–88. doi:10.1016/j.critrevonc.2018.03.005Yan, G., Zhang, Y., and Wang, J. (2017). Recent advances in the synthesis of aryl nitrilecompounds. Adv. Synthesis & Catal. 359 (23), 4068–4105. doi:10.1002/adsc.201700875Youssif, B. G., Gouda, A. M., Moustafa, A. H., Abdelhamid, A. A., Gomaa, H. A.,Kamal, I., et al. (2022). Design and synthesis of new triarylimidazole derivatives as dualinhibitors of BRAFV600E/p38α with potential antiproliferative activity. J. Mol. Struct.1253, 132218. doi:10.1016/j.molstruc.2021.132218Youssif, B. G., Mohamed, M. F., Al-Sanea, M. M., Moustafa, A. H., Abdelhamid,A. A., and Gomaa, H. A. (2019). Novel aryl carboximidamide and 3-aryl-1, 2, 4-oxadiazole analogues of naproxen as dual selective COX-2/15-LOX inhibitors:design, synthesis and docking studies. Bioorg. Chem. 85, 577–584. doi:10.1016/j.bioorg.2019.02.043Zhang, Q., Diao, Y., Wang, F., Fu, Y., Tang, F., You, Q., et al. (2013). Design anddiscovery of 4-anilinoquinazoline ureas as multikinase inhibitors targeting BRAF,VEGFR-2 and EGFR. MedChemComm 4 (6), 979–986. doi:10.1039/c3md00096fZhou, J., Jiang, X., He, S., Jiang, H., Feng, F., Liu, W., et al. (2019). Rational design ofmultitarget-directed ligands: strategies and emerging paradigms. J. Med. Chem. 62 (20),8881–8914. doi:10.1021/acs.jmedchem.9b00017Frontiers in Chemistry frontiersin.org17Mohamed et al. 10.3389/fchem.2024.1447618', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explai"}}},{"rowIdx":93,"cells":{"Researcher Name":{"kind":"string","value":"Bin Wu"},"ORCID":{"kind":"string","value":"0000-0002-8152-4011"},"Researcher Query Keyword":{"kind":"string","value":"Ion-Selective Membranes with Crown Ether-MOF Integration"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":94,"cells":{"Researcher Name":{"kind":"string","value":"Xingya Li"},"ORCID":{"kind":"string","value":"0000-0002-2824-6809"},"Researcher Query Keyword":{"kind":"string","value":"Ion-Selective Membranes with Crown Ether-MOF Integration"},"Research Full Paper":{"kind":"string","value":"{'Metal–Organic Framework–Based Ion–Selective Membranes': 'Title: Metal–Organic Framework–Based Ion–Selective Membranes\\\\nAbstract Inorganic contaminants, such as fluoride and arsenic are problematic inorganic contaminants due to major human health risk and relatively high levels of occurrence. Metal-organic frameworks, (MOFs) are a novel approach to adsorption of fluoride and arsenic that have a high surface area, versatile building blocks and numerous active sites. This article presents a comprehensive review on different types of MOFs for fluoride and arsenic removal along with dynamic breakthrough time and cost analysis. Performances of MOFs are based on a variety of synthesis method, notable among which is solvothermal synthesis. However, it is evident from all the research conducted that MOFs have poor yield compared to conventional adsorbents. But, their high adsorption capacity, tailored chemical structure and ionic uptake of fluoride and arsenic makes them more favourable option than the other adsorbents. Material price of different MOFs usually varies between 0.1-5 USD/gram, which is shown in this study. Keywords: Metal-organic framework, MOFs; adsorption; arsenic removal; fluoride removal; fixed bed sorption; cost analysis Accepted Preprint Manuscript2 Contents 1 Introduction 3 2 Contamination of groundwater by fluoride and arsenic and related side-effects 4 3 Conventional removal methods for the decontamination of fluoride and arsenic ions from aqueous environment 6 3.1 Adsorption 8 4 Superiority of MOFs in fluoride and arsenic adsorption 13 5 Preparation of different varieties of MOFs 15 5.1 Structure 15 5.2 Synthesis 16 5.3 Classifications 17 6 Different MOF based adsorbents for fluoride and arsenic uptake 17 6.1 Uptake mechanism 18 6.2 Usage in fixed bed column study 20 6.3 Cost analysis of MOFs used in fluoride and arsenic removal 21 7 Conclusion 22 8 Nomenclature 22 Accepted Preprint Manuscript3 1. Introduction Water is an important concern, globally. Water stress has increased because of heavy agricultural and industrial demand, and changes in hydrologic cycles driven by climate change [1]. It is estimated that agricultural and energy water demand will continue to increase between 60% - 80% for the agricultural and energy production activities by 2025, while the rise in human population is expected in the range of 22–32% by 2050 [2, 3]. Pollution of surface and groundwater remains a problem as well. Uncontrolled release of chemicals to the environment such as heavy metals, organics, inorganics, and fertilizers have deteriorated water quality [4, 5]. Some inorganic contaminants are persistent and usually well removed by conventional water treatment methods. Amongst these inorganic contaminants, fluoride (existing as F–) and arsenic (existing as arsenite, AsO33– and arsenate AsO45-) are harmful and extensively found in groundwater. Pollution related to fluoride and arsenic occurs via geologic weathering and anthropogenic activities [6, 7]. Fluorides contamination effects an estimated 62 million people, and 300 million are impacted by arsenic contaminated water [8]. Continuous consumption of fluoride and arsenic causes health problems, such as fluorosis, keratosis of hands and feet, hyperpigmentation, cancerous outgrowths in kidney, lungs, bladder, skin and liver [9, 10]. Therefore, it is necessary in many situations to remove fluoride and arsenic to provide safe drinking water. Mitigative options for arsenic and fluoride treatment exist including electrocoagulation [11], precipitation [12], floatation [13], anion exchange [14], filtration with nanofiltration (NF) and reverse osmosis (RO) membranes [15, 16], electro-dialysis [17] and adsorption [18]. However, these methods have some potential drawbacks. For example, precipitation method produces toxic by-products, requiring removal prior to final adsorption [19], flotation requires substantial quantity of flocculating agents [20], regeneration in ion exchange carries brine disposal issues [21, 22] and fouling problems exist in NF and RO membranes [16, 23]. Accepted Preprint Manuscript4 Adsorption with porous materials is another option for removal of arsenic and fluoride from potable water, offering relative ease in operation but with competition from other co-occurring adsorbates [24]. Metal-organic frameworks (MOFs) offer a potential alternative method from adsorption of fluoride and arsenic. MOFs have high surface area with abundant active sites for adsorption, as compared to other conventional adsorbents [25]. They also possess ordered crystalline structure, made of organic-inorganic hybrid networks [26]. MOFs are also known to exhibit superior physicochemical characteristics with varied adsorption applications, such as, uptake of hydrogen [27], carbon dioxide [28] and other toxic gases [29]. Due to their promising potential in these areas, researchers have started to investigate the capability of MOFs in removal of metal ions [30], as well as toxic dyes [31], herbicides [32] and humic acid [33] from groundwater. In general, MOFs have an adaptable and porous structure, allowing diffusion of ionic contaminants into their bulk structure [34]. Various studies reveal that certain MOFs show excellent uptake capacities. MOFs have also been evaluated for fluoride and arsenic (As(III) and As(V)) removal from groundwater. It was observed that MOFs exhibit high uptake capacity for fluoride (32 mg/g) and arsenic (12 mg/g) [35, 36]. The adsorptive properties of MOFs have further been enhanced by various chemical and structural modifications. The aim of this review is to provide an in-depth analysis about the latest development of MOFs related to fluoride and arsenic removal from groundwater. This work also presents the adsorption kinetics of selected MOFs in the dynamic purification of fluoride and arsenic contaminated streams via an empirical model along with its cost estimation. 2. Contamination of groundwater by fluoride and arsenic and related side-effects A large section of the global population depends on groundwater as their primary source of water [37]. However, groundwater contamination by fluoride and arsenic leads to chronic poisonings and endemic diseases in some of these individuals. Mobilization of these two toxic Accepted Preprint Manuscript5 contaminants under natural conditions is one of the main reasons for groundwater contamination [38]. The primary anthropogenic cause of fluoride pollution is mining and fertilizers production, especially, phosphate fertilizers. The element fluorine usually exists as fluorides in nature. It occurs mainly as magnesium fluoride (sellaite – MgF2), calcium fluoride (fluorspar – CaF2), sodium hexafluoroaluminate (cryolite – Na3AlF6) and fluoropartite (Ca₅(PO₄)₃F) [39]. Depending on the pH of the medium, fluoride ion can form soluble complexes with polyvalent cations such as aluminium (Al3+), magnesium (Mg2+), calcium (Ca2+) and iron (Fe3+) [40]. Consumption of water containing excess of fluoride has profound negative effects on hydroxyapatite, (Ca5(PO4)3OH) which is a mineral constituent of teeth and bones. Fluoropartite enhances density and hardness of teeth and bones, thereby making it brittle [10]. This condition is termed as fluorosis. Dental fluorosis usually occurs in children when they chronically ingest fluoride-contaminated water [41]. However, skeletal fluorosis results in crippled anatomical structures mostly in adults. Apart from dental and skeletal fluorosis, fluoride exposure affects structure and functioning of skeletal muscle, brain and spinal cord and alters metabolism of essential nutrients. This leads to physical conditions such as hyperkalaemia, hypocalcaemia, hypomagnesemia and hypophosphatemia [42, 43]. Considerable range of 0.5-1.5 mg/L fluoride in water is good for the health of teeth and bones, while the maximum limiting concentration of fluoride in water, suggested by the WHO, is 1.5 mg/L [44]. Maximum contaminant level (MCL) of fluoride enforced by the USEPA is 4.0 mg/L [45]. Arsenic is mobilized by volcanic emissions, biological activities, natural geochemical reactions, and anthropogenic causes. The dissolved and suspended forms of arsenic released by soil erosions and leaching, contribute to 612×108 and 2380×108 g/year, respectively [46]. Arsenic usually exists as arsenite (AsO33−) and arsenate (AsO45−) forms in natural waters, which are commonly referred to as As(III) and As (V). As(OH)3 undergoes protonation (pKa = Accepted Preprint Manuscript6 9.2); hence As(III) exists mainly in the form of uncharged arsenous acid (H3AsO3) in contaminated stream. On the other hand, As(V) exists as oxyanions in the forms of arsenic acid, H2AsO4- (pKa - 2.3) and HAsO42- (pKa - 7.1) in oxidizing waters (pH ranging from 6.5-8.5). Pentavalent species are stable in aerobic environments, while the trivalent arsenite are present in moderately reducing anaerobic environments [47]. Long term exposure of arsenic in drinking water, causes cancer of the skin, lung, bladder, and kidney. These are augmented with pigmentation changes, hyperkeratosis, myocardial infection, hypertension, diabetes, neurological disorders, muscular weakness, loss of appetite and nausea [9, 48]. Chronic exposure to arsenic results in Bowen’s disease, characterized by inflammation of stomach and intestines, fatigue, kidney degeneration, bone-marrow degeneration, cirrhosis of the liver, and severe dermatitis. It is observed that arsenic shares many properties with tumour promoters, by activating transcription factors, intracellular signal transduction and changing expression of genes that are involved in promoting cell growth [49, 50]. A WHO provisional guideline of 10 µg/L of arsenic has been adopted as the drinking water standard in many areas. However, some countries have retained the earlier standard of 50 µg/L [51]. The USEPA MCL for arsenic is 0.010 mg/L [52]. 3. Conventional removal methods for the decontamination of fluoride and arsenic ions from aqueous environment Techniques such as, precipitation, coagulation or electrocoagulation, membrane filtration and adsorption are mostly used to treat water contaminated with fluoride and arsenic. Discussions related to each removal method along with their advantages and disadvantages are described in the following section. A summary of different methods for fluoride and arsenic uptake with the specific agent used in the process, has been presented in Table 1 [Table 1 near here]. Accepted Preprint Manuscript7 Studies such as a two-column limestone reactor [13], crushed limestone [53], co-precipitation with calcium enhanced ferric hydroxides [54] and calcium phosphate hydroxyapatite [55] were reported for fluoride and arsenic removal by precipitation method. It was concluded from these studies that this method was relatively easy in operation but its major disadvantages were initial cost and sludge disposal. Similarly, both electrocoagulation and chemical coagulation have been examined for fluoride and arsenic removal. For example, defluoridation was investigated using a continuous flow electrocoagulation reactor [56], parallel-plate electro-coagulation process [57] and a natural coagulant Moringa oleifera (MO) [58]. While arsenic removal was studied in a mine drainage system using coarse calcite coated ferric sulphate [59], electrocoagulation using mild and stainless plates [60] and aluminium alloy and stainless-steel electrodes [61]. The studies suggested the process to be both eco-friendly and cost-effective option for municipal and industrial water treatment, but requires continuous addition of substantial quantities of coagulants and further sludge treatment. Membrane separation techniques such as reverse osmosis (RO), Nanofiltration (NF) and electro-dialysis (ED) are commonly used for fluoride and arsenic removal. The feasibility of applying RO membranes to process electronic industrial effluent with a load of one kg/day of fluoride [62], portable ultra-low-pressure RO system (ULPRO) to remove fluoride and total dissolved solids (TDS) from coalbed methane (CBM)-produced water [63] and removal of arsenic using seawater high rejection (SWHR) and brackish water (BW-30) membranes [64] have been reported. In these studies, main focus was on RO process. In another study fluoride removal from a high fluoride stream was studied using negatively charged commercial thin-film composite (TFC) membranes [65], flat sheet crossflow NF module was utilised for fluoride removal from contaminated groundwater [66] and arsenic removal by commercial TFC membrane [67, 68] were reported. Fluoride removal from aqueous solution was studied by electro-dialysis, using TS-l-l0 (Tokuyama) electro-dialysis equipment [69], SB-6407 anion Accepted Preprint Manuscript8 exchange membrane [70], corning P1 electro-dialysis pilot equipment [71], while for arsenic removal a laboratory-scale electrodialyzer [72] and BEL-500 stack [73] have been reported. In these studies, removal efficiency was better over precipitation and coagulation, but the major disadvantage of these processes is fouling and high cost. 3.1. Adsorption Precipitation, coagulation, electrocoagulation and membrane separation successfully decrease fluoride and arsenic concentrations to acceptable levels; however, they have certain limitations. For example, precipitation and coagulation are economical, but they are known to produce a large amount of sludge that needs further treatment [19, 20]. Membrane separation processes are usually energy intensive, compared to conventional treatment [16, 23]. Adsorption offers critical advantages, such as robustness, lower operating cost, greater efficiency, and, most importantly, the possibility of using versatile adsorbent materials [24]. For example, activated carbons include a versatile range of carbonaceous adsorbents, which can be made to increase surface area and porosity. The use of activated carbon dates back to ancient times, while its current usage for water treatment has begun in the second half of the 20th century [74]. Activated carbon is known to have strong affinity to these organic and inorganic contaminants, even at lower concentrations, thereby making it a high performance and low energy alternative. Adsorption has gained full-scale acceptance in fluoride and arsenic removal from industrial wastewater and groundwater. It is well known that adsorption is a surface phenomenon, hence surface ions and pH play a vital role in the process. Adsorption mechanism [75] of fluoride ions on solid particles can be explained with the following steps: (1) mass transfer of fluoride ions from bulk stream to external surface of the adsorbent, known as external mass transfer; (2) adsorption of fluoride ions on the external surface of the particle; (3) transfer of the adsorbed fluoride ions to the internal surfaces of the porous material, i.e., intraparticle diffusion or possible exchange of Accepted Preprint Mauscript9 adsorbed fluoride ions with structural elements inside the adsorbent. Removal of fluoride has been studied using different adsorbents. One such example is porous magnesium oxide (MgO) nanoplates [76]. pH had a minimum effect in the range 2–11. However, when the pH exceeded 12, fluoride removal rate was decreased because hydroxyl ions compete with fluoride ions for adsorption sites in alkaline pH, resulting in the reduction of fluoride removal percentage. A Fourier-transform infrared (FTIR) spectroscopy study suggested that total amount of carbonates on MgO decreased during fluoride adsorption process. It is well accepted that fluoride ions can be adsorbed on MgO surface via surface hydroxyl and carbonate exchange. Similar results were obtained from XPS, confirming that coexistence of bicarbonate and carbonate ions has an enormous impact on adsorption performance. The adsorption mechanism for fluoride removal was also investigated using zirconium oxide (ZrO2) mesoporous fibers [77]. It was observed that pH substantially affected adsorption capacity. When the pH of the solution was below isoelectric point (pH = 4) of ZrO2, hydroxyl groups on the surface of the fibers were protonated, resulting an increase in active sites and reinforcing interaction between the adsorbent and fluoride. This contrasts with the surface of ZrO2 fibers carrying negative charges at high pH, thereby repelling fluoride. Additionally, there was a decline in fluoride uptake capacity when the pH was reduced to 2, which was due to sparingly soluble hydrofluoric acid formation. FTIR and XPS studies suggested an ion-exchange mechanism involved in fluoride adsorption, as shown below in Equation 1: -Zr-OH(S) + F -(L) → -Zr-F(S) + OH- (L) (1) Yadav et al., also studied removal of fluoride from aqueous solution by three low-cost agricultural biomass-based adsorbents: wheat straw raw (WSR), sawdust raw (SDR) and activated bagasse carbon (ABC) [78]. It was observed that pH and functional groups on the bio-sorbent played a major role in fluoride adsorption [79]. Maximum biosorption values were observed at acidic pH, where overall surface charges are likely positive, enabling binding of Accepted Preprint Manuscript10 negatively charged fluoride ions. At lower pH, surface of adsorbent turned out to be positively charged while relative sorption inhibition was observed at basic pH attributing to increase in hydroxyl ion and the formation of aqua complexes. From the kinetic study, the dual nature of the adsorption isotherms support that the initial curve is due to the boundary layer diffusion, while the final linear portion is due to intraparticle diffusion. This indicated that the mechanism of fluoride removal was complex. Hence, both surface adsorption and intraparticle diffusion contributed to the rate determining step. Defluoridation of water was also investigated using pine wood and pine bark-based biochar and it was prepared by pyrolysis at 4000C and 450℃ in a reactor [80]. As-synthesized chars were characterized and used for defluoridation of acidic groundwater. Swelling of the chars was observed attributable to increased oxygen percentage which subsequently opened new internal pores within the sample. This lead to diffusion of fluoride ions into the subsurface, thereby promoting further adsorption. Ion exchange was observed as the mode of uptake. The authors further concluded that these chars could have been tested for activation and enhancement of surface area for increased uptake, as they had greater ability to remove higher fluoride content than activated carbon. Similarly, activated carbon were prepared from bark of Morinda tinctoria coated with aluminium hydroxide [81]. Industrial bone char [82] and bone char from Pleco fish (Pterygoplichthys spp.) [83] were also used for fluoride uptake. Similar studies were performed with metal activated carbon [84], rice husk [85], bael shell [86], calcium chloride modified Crocus sativus leaves [87]. These are all natural products which were physically and chemically activated. Tri-metal oxide and rare earth minerals were also used for activation. For example, Mg-Mn-Zr impregnated activated carbon [88], cerium impregnated activated carbon-based novel composite [89] have been reported. These studies revealed that adsorption was primary mechanism for uptake, where surface charges, pH, ion exchange, intraparticle diffusion, electrostatic attraction/repulsion were important factors, as explained above. Accepted Preprint Manuscript11 Removal of arsenic was also conducted by different adsorbents. For example, arsenic removal by feldspar was studied, concluding that both electrostatic forces and chemical interactions were the adsorption mechanisms [90]. The positively charged surface of the mineral is obtained when pH is lower than the point of zero charge (pHPZC) of the sorbent, leading to more favourable adsorption of As(V). Surface ionization reactions occur, depending on pH of the solution containing the mineral. Surface protonation (Equation 2) is promoted in acidic medium, whereas, deprotonation reaction (Equation 3) is increased under basic conditions, as shown below: Als – OH + H+ ↔ Als – OH2+ (2) Als – OH ↔ Als – O- + H+ (3) where Als – OH is neutral aluminol, Als – OH2+ is protonated aluminol, Als – O- is hydrolysed aluminol and H+ is proton or hydrogen ion. Both the positively charged surface of solid (aluminol active sites present as ≡Al-OH+) and the predominant speciation of H2AsO4− are favourable for As(V) uptake at acidic condition. However, adsorption of monodentate non-ionized arsenite (As(III)) was contrary to As(V), occurring only through a ligand exchange reaction and, most favourably, onto the non-ionized surface functional groups. The authors concluded that coulombic interaction and solution pH is practically insignificant for As(III) adsorption as compared to As(V). Arsenic removal was also studied using iron oxide modified thermally produced cigarette soot activated carbon (CSAC) i.e. (Fe3O4/CSAC) adsorbent [91]. In this study, it was observed that arsenic adsorption was highly pH dependent. pHpzc of the adsorbent decreased as arsenic anions make the surface charge more negative. It was evident from FTIR studies that removal of hydroxyl functional groups from adsorbent surface took place, generating complexes such as As(III) – Fe3O4/CSAC and As(V) – Fe3O4/CSAC. This confirmed a possible mechanism of monodentate and bidentate ligand exchange for arsenic. In case of arsenate adsorption, protonated hydroxyl Accepted Preprint Manuscript12 groups electrostatically attracted As(V) species (i.e., H2AsO4−) in acidic solution, forming monodentate and bidentate complexes, between the M–OH groups. Arsenite, being present in uncharged form in neutral pH (pH 7), gets adsorbed on Fe3O4/CSAC surfaces by forming monodentate complexes. Arsenic removal was also investigated using magnetic biochar [92]. Results demonstrated that adsorption of As(V) on the biochar/γ-Fe2O3 composite was relatively fast and reached equilibrium within four hours. This rapid kinetics suggested that biochar might play an important role in the dispersion of γ-Fe2O3 particles, which increased the surface area of particles and active sites of metal oxides. A similar kinetic study also suggested that adsorption of As(V) to metal oxide surfaces is mainly through surface complexation reactions and can be described by one and two-site models. Monodentate and bidentate As(V) adsorption reactions can be written as follows: Monodentate: SOH + H3AsO4 ↔ SHmAsO4(m – 2) + (2 – m)H+ +H2O (4) Bidentate: S(OH)2 + H3AsO4 ↔ SHnAsO4(n – 1) + (1 – n) H+ +H2O (5) where S is γ-Fe2O3, m having a value of 0, 1 or 2 and n can be 0 or 1. These adsorption reactions were observed to be monolayer and site-limited and thus confirmed to Langmuir adsorption theory. A microporous activated carbon was prepared for arsenic removal from water [93]. Surface functionality of iron loaded activated carbon (FCAC) is highly pH dependent and affects arsenic adsorption on its surface. Positive charge on the surface is converted to negative charge by deprotonation due to abundance of hydroxyl ion at higher pH. On the other hand, negatively charged species faces electrostatic repulsion, which is attributed to its lower adsorption, in addition to higher number of organic functional groups, which are randomly distributed on the chemical activated carbon (CAC) surface than that of physical activated carbon (PAC). It was also observed that CAC adsorbed iron species, thereby decreasing pore size. In arsenic uptake, micropores and surface functional groups played significant roles, as CAC performed much better than PAC. However, none of them produced Accepted Preprint Manuscript13 the desired separation of arsenic from water. In contrast, FCAC could achieve the desired uptake of arsenic due to iron, which produced a strong affinity between surface functional groups and arsenic species. Similarly, other adsorbents such as Sargassum muticum coated with iron-oxy(hydroxides) [94], iron hydroxide/manganese dioxide doped straw activated carbon [95], Perilla leaf biochar [96], Japanese oak wood biochar [97] are few examples of natural adsorbents that have been physically and chemically activated. In some studies, rare earth metals were used for activation. For example, cerium oxide modified activated carbon [98], CeO2/Fe2O3/graphene nanocomposite [99] and Halloysite-CeOx (x = 1.5–2.0) nanocomposite [100] have been reported. Similar studies were performed using iron for impregnation. For example, magnetic gelatin modified biochar [101], Fe3O4@Al2O3@ Zn-Fe LDH (LDH – layered double hydroxides) [102] and iron-incorporated activated carbon from biomass mixture [103] have been reported. Modifications were also done using mesoporous bismuth-impregnated aluminum oxide [104], aluminum-based adsorbent and coal mine drainage sludge coated polyurethane [105] and covalent triazine framework encapsulated γ- Fe2O3 nanoparticles [106]. These studies revealed that adsorption mechanism was mainly dominated by monodentate and bidentate complex formation and electrostatic attraction or repulsion. Table 2 summarises list of adsorbents for fluoride and arsenic uptake with their respective adsorption capacities and mechanism. [Table 2 near here] 4. Superiority of MOFs in fluoride and arsenic adsorption As outlined in Section 3, various approaches have been adopted for defluoridation and arsenic removal from groundwater. Precipitation [13, 55], electrocoagulation [57, 60], membrane separation [62, 69] and adsorption techniques [76, 85] are some of the most common and efficient methods for the same. Recently, MOFs have drawn increased research interest because of their unique properties. They are made of organic and inorganic material and possess crystalline structure, have a large internal surface area (more than 6000 m2/g), as well Accepted Preprint Mauscript14 as high thermal and chemical stability with high porosity (greater than 80%) [26]. These MOFs consist of a positively charged metal ion surrounded by an organic linker, forming a cage-like hybrid structure. MOFs have some inherent advantages, compared to other porous materials in terms of atomic-level structural uniformity, tuneable porosity, uniform pore structures, flexible network topology and other chemical and geometric identities [25, 26]. Researchers have started to explore the use of MOFs in water treatment and have observed some promising results [30]. Several research articles have been published (approximately 1500) related to MOFs for water treatment. Few of them have been cited here as an example for examining MOFs to mitigate metal ions [30], dyes [31] and herbicides [32]. These studies suggest that different MOFs are stable in various aqueous solutions, and exhibit relatively high contaminant removal. Recently, MOFs have also been examined for fluoride and arsenic removal [35, 36]. Table 3 summarizes the adsorption capacities of various MOFs that are utilized for fluoride and arsenic uptake. It is evident that MOFs exhibit high adsorption capacity for these ions, compared to other adsorbents. For example, alumina treated activated carbon has a fluoride adsorption capacity of 4.5 mg/g [84], while that of aluminium fumarate MOF is 600 mg/g [108]. This MOF has similar chemical composition to the conventional adsorbent, but a much greater defluoridation capacity is due to its large surface area (1156 m2/g). With respect to arsenic sorption, zirconium-based nanoparticle doped activated carbon fibers have a reported uptake capacity of 21 mg/g [101], while that of zirconium MOF is 303 mg/g [109]. However, surface area of the former is observed to be higher. Zirconium based activated carbon fibers have surface area of 1409 m2/g, while that of zirconium MOF is 570 m2/g. The authors cited that high uptake capacity for MOF can be then attributed probably to its internal cage structure. Constricted chains in the cage structure perhaps results in lower surface area. But this cage structure provides more space to uptake of fluoride than conventional adsorbents. Detailed discussion regarding their preparation routes, Accepted Preprint Manuscript15 characterization, uptake capacities and mechanism have been performed in the next sections, which highlight superior performance of MOFs as compared to traditional sorbents. 5. Preparation of different varieties of MOFs 5.1. Structure MOFs are composed of different unit types: (1) secondary building units (SBUs), which are essentially a cluster of metal ions, and (2) organic linkages between subsequent SBUs that gives rise to highly porous crystalline structure. The organic units are typically mono, di, tri or tetravalent ligands. The choice of metal and linker decide the structure and properties of MOF. For instance, the metal coordination preference impacts size and shape of pores. This coordination also controls the number of ligands binding to the metal along with its orientation. SBUs are attached by the bridging ligands. Typical bridging ligands are di and tricarboxylic acids. Examples are benzene-1,4-dicarboxylic acid (BDC) or terephthalic acid and biphenyl-4,4\\\\'-dicarboxylic acid (BPDC) [110]. [Figure 1 near here] Figure 1 shows SBU (metal node) and organic linker used for synthesis of IRMOF – 1 (IR - Isoreticular) and HKUST-1(Hong Kong University of Science and Technology) [111]. The combination of these structures results in a large number of possible arrangements. MOFs can also be synthesized using same SBU, but different organic linkers [112], as shown in Figure 2. [Figure 2 near here]. Diverse pore shapes can be obtained depending on the linker. Figure 3 shows representative SBUs and organic linkers used for synthesis of MOFs [113]. SBUs has served as an organizing concept for classification of MOFs structures into their underlying topology. They are also essential to design the directionality of MOFs construction and achieve robust frameworks. It also ensures thermodynamic, mechanical and, architectural stability, originating from strong directional bonds, thereby locking down the position of metal centres [114] [Figure 3 near here] Accepted Preprint Manuscript16 5.2. Synthesis Figure 4 summarises commonly used synthesis methods for MOFs. Usually, MOFs are prepared by mixing metal salts, organic ligands and solvent for specific time duration (usually between 12 and 48 hours) [25]. The product formed from the reaction is then filtered and dried. A common synthesis method of MOFs is hydrothermal or solvothermal synthesis [115]. Generally, high-solubility organic solvents, such as dimethyl-formamide, ethanol, acetonitrile, methanol and acetone are used in solvothermal reactions. Mixtures of various solvents can be used to avoid problems, related to solubility of initial reagents. Solvothermal synthesis is generally performed in Teflon-lined autoclaves at temperatures higher than 130℃ [115]. [Figure 4 near here]. Other synthesis routes such as microwave-assisted, electrochemical, sonochemical, mechanochemical and spray-drying have been a focus of recent research. Solid state-based synthesis method of MOFs also exists [116]. No external energy supply is required in this method. This route collectively uses different solvents to increase solubility of the reagents and accelerate synthesis by rapid evaporation of solvents. These methods can synthesize a greater quantity of MOFs in a smaller period of time [117]. Thermodynamic and activation energy of the synthesizing reaction is based on the redox potential, reactivity and solubility of the solvent. Microwave-assisted synthesis provides advantage of short synthesis time, porous texture, reduced particle size, better morphology and high crystallinity, compared to solvothermal method. However, this method is solvent sensitive as it involves interaction of the solvent with electromagnetic and electrical waves [118]. Mechanochemical synthesis works on mechanical agitation and collision between substances. This mechanochemical approaches offers the advantages of process efficiency, simplicity, no solvent usage and low energy consumption [119]. Intensive ultrasonic radiation (20 kHz–10 MHz) is applied in sonochemical synthesis producing cavitation–the generation of small bubbles within liquid Accepted Preprint Manuscript17 phase. Produced bubbles collapse momentarily creating localized moments of high temperature and pressure. These ephemeral areas generate homogenous nucleation centres that decrease crystallization time as compared to more conventional solvothermal methods [120]. Electrochemical synthesis produces MOFs using thin films over surfaces at mild temperatures, reducing the effect of film cracking of metal salts in the solution, while offering its continuous production. Electrochemical synthesis is important from industrial process perspective as it offers continuous production of MOFs [121]. Figure 5 summarises such methods with reaction temperatures and final reaction products in MOF synthesis [122]. [Figure 5 near here]. 5.3. Classifications Crystal engineering of MOFs not only relies on permanent porosity, but also on other factors post-modification including reversible structural transformation, and framework integrity. In this context, an attributive classification of MOFs helps in understanding their structure. As depicted in Figure 6, first generation MOFs have only non-permanent porosity because of inseparable host–guest dependence. This phenomenon has often been observed in MOFs containing charged frameworks, with pores filled by counter anions. Comparatively, second generation MOFs possess stable and robust porosity against guest removal such as in neutral and zeolite-like MOFs. Third generation MOFs display framework flexibility and dynamics, and are able to respond to guest exchange or external stimuli. Fourth generation MOFs are correlated to recently developed post-synthetic modifications (PSM) and may be broadly defined as post-processing MOFs. They can maintain underlying topology and structural integrity towards various PSM [123]. [Figure 6 near here] 6. Different MOF based adsorbents for fluoride and arsenic uptake Different studies have been performed to explore the utility of MOFs as an efficient adsorbent for water treatment. For example, MOFs have been utilized for heavy metals removal [30] and dyes [31]. Different aspects including the uptake mechanism usage in fixed bed Accepted Preprint Manusc ipt18 columns for dynamic operations and the manufacturing cost have been discussed in the following sections. 6.1. Uptake mechanism Different MOFs and their respective uptake mechanisms are summarised in Table 3. It is usually observed that intraparticle diffusion [124] and the specific nature of surface ions [125] govern the adsorption mechanism in MOFs. Figure 7 depicts the possible uptake mechanisms for adsorptive removal [126]. Defluoridation mechanism was studied using water stable MOFs [127]. Eleven water-stable MOFs, MIL-53 (containing, iron, chromium and aluminium, MIL - Matériaux de l′Institut Lavoisier), MIL-68 (containing aluminium), aluminium based MOFs (CAU-1, CAU-6), UiO-66 (zirconium and hydrofluoric acid) and ZIFs-7, 8, 9 (zeolitic imidazole framework) were considered in this case. It was concluded from performed characterizations that certain best practices should be observed to design MOFs with better stability in fluoride spiked aqueous solution. For example, a relatively inert metal can be better choice for MOFs with the same topology [127]. In the same study, pore topology was considered for materials with same metal cluster. It was also suggested that coordination number of metals should be high and appropriate hydrophobicity can lead to good stability of MOFs. The authors used UiO-66(Zr) for removal of fluoride from water because of higher adsorption capacity. Defluoridation capacity of MOFs are dependent on concentration of other co-existing ions in the solution. Also, increasing the number of hydroxyl groups is an efficient strategy to improve MOF performance. From the kinetic study, it was observed that mechanism of fluoride adsorption on UiO-66(Zr) is complex and both surface adsorption and intraparticle diffusion contribute to rate-determining step [127]. It was also suggested that fluoride ions are adsorbed onto UiO-66(Zr), as its pore size is larger than ionic radius of fluoride. Two lanthanide-based MOFs were prepared and investigated for removal of fluoride from water [128]. Adsorption was facilitated at lower pH (3-7), but it dropped drastically after pH 8. At Accepted Preprint Manuscript19 lower pH, lanthanide-based MOFs are positively charged and nucleophilic replacement of hydroxyl ions by fluoride essentially uninhibited. However, many MOFs decompose in a very acid medium, and at pH ≤ 2 most MOFs have negligible adsorption effectiveness. Nucleophilic replacements are not preferred at higher pH because of abundance of hydroxyl ions, which competes with fluoride for the active sites in the adsorbent. Therefore, it is apparent that adsorption of fluoride from aqueous media is best carried out at pH ranging from 3 to 7. Similar to previous studies, zirconium-MOF [129], UiO-66-NH2 [130], MOF – 801 [131], calcium fumarate (CaFu) [132] and UiO-66-amine [133] were mostly dominated by surface ion exchange and intraparticle diffusion mechanism for fluoride adsorption. [Table 3 near here] The removal of arsenic from aqueous solution using ZIF-8 MOF [134] was studied. The pHZPC of ZIF-8 MOF was 9.6 and this is stable at neutral and basic conditions. In acidic conditions, ZIF-8 MOF dissolves in the stream which dramatically hinders adsorption of arsenic. An intraparticle diffusion model was used to analyse rate-controlling step, based on the kinetics data in order to identify arsenic transport process across ZIF-8 MOF. The authors concluded from kinetic studies that adsorption rates were mainly controlled by pore diffusion rather than mass transfer through boundary layer. Zn-MOF-74 was also utilized for ultra-trace quantity arsenic removal [135]. Maximum removal of As(V) was achieved at pH 6-8. Adsorption mechanism was due to electrostatic interaction or chemical reaction between the arsenate and functional groups on the surface of Zn-MOF-74. It was concluded that adsorption of arsenate onto this adsorbent was due to chemical interactions, leading to substitution of water molecules inside pore channels. It was also observed that intraparticle diffusion is the rate controlling step during this process. Similarly, other MOFs, such as, AUBM-1 (AUBM: American University of Beirut Materials) [136], NH2 -MIL-88(Fe) [137], MIL-53(Fe) [138], MIL-101(OH)3 and MIL-101(OH) [139], MOF-808 [140], MIL-53(Al) [141], ZIF-8 [142], UiO-66 incorporated thin nanocomposite membrane [143], Fe3O4@MIL-101 [144], ZIF-8 Accepted Preprint Manuscript20 (cetyltrimethylammonium bromide (CTAB) and amino acid L-histidine (His) as co-templates) [145] and BUC – 17 [146] were reported for arsenic adsorption. These MOFs are influenced by electrostatic, acid-base and coordination interactions, along with hydrogen bonding. 6.2 Usage in fixed bed column study It is known that data obtained from batch studies are usually limited to laboratory scale. Complication are incurred in employing MOFs at full-scale. Column studies become necessary to obtain data for design of continuous flow sorption model [147]. Various MOFs have higher adsorption capacity compared to conventional adsorbents. Hence, it becomes important to test the MOFs in a continuous flow operation. Most of the research articles concerning MOFs for water treatment did not portray fixed bed column studies. Hence, in this article, we have estimated the breakthrough time for continuous flow operation by using sorption data from batch studies and few assumptions. These assumptions are (a) packed bed containing 5 kg MOF as adsorbent; (b) packed bed can treat 40 L/day of feed solution and (c) feed solution have fluoride and arsenic concentration (Co) of 10 mg/L and 5 mg/L, respectively. An empirical model, i.e., Thomas model [148] has been used to estimate breakthrough time of these MOFs in fixed-bed adsorption column. Linearized form of this model can be described by the following expression, ln(%&%\\\\'− 1) = *+,-./0−\\\\t𝑘𝑇𝐻𝐶𝑜𝑡 (6) where, kTH is Thomas model constant (mL/min mg), qe is adsorption capacity (mg/g), Co is effluent concentration (mg/L), m is adsorbent dosage (g), Q is feed flowrate (mL/min), t is breakthrough time (min). Thomas model constant, kTH has values in the range 0.08-0.03 mL/min mg for most adsorbents in reported articles [149]. An average value of 0.051 mL/min mg for Thomas model constant, kTH has been chosen in this article. Breakthrough time have been calculated for different adsorbents, related to arsenic and fluoride removal. These values are presented in the Table 4. It is observed in Table 4 that calcium fumarate MOF [132] will Acceptd Preprint Manuscript21 show maximum breakthrough time (25 hours) to produce filtrate, having fluoride concentration lesser than allowable limit (1.5 mg/L), while MOF – 801 [131] attains the lowest one (6 hours). Similarly, maximum breakthrough time having arsenic concentration lesser than 10 µg/L will be attained for Zr-MOF (UiO-66) (44 hours) [109] while the lowest one by MIL– 53(Fe) (6 hours) [138]. These studies suggest that the prepared MOFs can be utilized for continuous flow operations, which can later be scaled up to a community-based filter. [Table 4 near here]. 6.3. Cost analysis of MOFs used in fluoride and arsenic removal Cost analysis of MOFs is necessary to assess the financial costs and/or benefits resulting from full-scale adaptation. A detailed cost analysis of MOFs is provided in Table 4. Table 4 includes results of a lanthanide-based MOF utilized for defluoridation, which cost approximate 1 USD/gram to manufacture [128]. MOF cost is inherently material dependent. For example, the cost of zirconium MOF [129] is observed to be an order of magnitude lower than the lanthanide-based example (0.1 USD/gram vs. 1 USD/gram). In the arsenic removal context, the cost of BUC–17 [146] MOF is the highest (5 USD/gram), while zirconium MOF [109] is the lowest (0.2 USD/gram). It is usually observed that the range of the manufacturing cost for one gram of MOF lies within 0.1 USD-5 USD. One can assume a median value of 2.5 USD per gram of MOF as its manufacturing cost, while the cost of other components to fabricate a filter is 45 USD. Hence, if 50 kg of a generic MOF is required to produce safe drinking water, then the total cost of the medium and the filter accessories can be calculated to be as 170 USD (rounded of to 150 USD). Most of the commercially available filters have a price ranging from 200-250 USD, where they claim to remove arsenic and fluoride from drinking water [143]. Therefore, based on the estimated cost, MOF based filter seems a frugal option. Accepted Preprint Mauscript22 7. Conclusions The current review gives a broader picture of water treatment via MOFs, focused on removal of fluoride and arsenic. Different remedial measures have been adopted to remove these persistent pollutants from water. However, these methods have high operating cost or they produce toxic sludge, which requires further treatment. On the other hand, MOFs are novel adsorbents, which have high uptake capacity for fluoride and arsenic and they can counter the disadvantages of conventional adsorbents to a great extent. Synthesis routes and building blocks can be altered, giving rise to versatile class of MOFs. However, MOFs remain restricted to lab-scale usage and are not explored for continuous flow sorption. MOFs based filter can be a cost-effective option in comparison to conventional filters, as shown in this study. However, low yield percentage of MOFs can be a blockade in their path for commercialization. The article also presents the breakthrough times of these MOFs in fixed bed adsorption columns, based on empirical mathematical model. Manufacturing cost of different MOFs along with filter accessories, have been estimated and it shows that they can be hosted an alternative option to large scale water treatment, especially related to fluoride and arsenic removal. 8. Nomenclature Abbreviations 4 – NP 4-nitrophenol ABC activated bagasse carbon AUBM American University of Beirut Materials BDC terepthalic acid/benzene-1,4-dicarboxylic acid BET Brunauer–Emmett–Teller Accepted Preprint Manuscript23 BIB bis-1H-imidazol-1-yl-methyl benzene BPY 4,4′-bipyridine BUT Beijing University of Technology BW - 30 brackish water - 30 CBM coalbed methane CNT carbon nanotube CP cycling properties CSAC cigarette soot activated carbon CTAB cetyltrimethylammonium bromide CV cyclic voltammetry DD desalination degree DPNI N, N-di-(4-pyridyl)-1,4,5,8-naphthalene tetra-carboxy-di-imide DSLF dual site Langmuir- Freundlich ED electrodialysis EIS electrochemical impedance spectroscopy FCAC iron loaded activated carbon FS flocculation – sedimentation FTIR Fourier-transform infrared spectroscopy H2BPDC 4,4′-biphenyl-dicarboxylic acid Accepted Preprint Manuscript24 H3TATB 4,4,4-s-triazine-2,4,6-triyl-tribenzoic acid HKUST -1 Hong Kong University of Science and Technology HRT hydraulic retention time IAST ideal adsorption solution theory IRMOF isoreticular metal organic framework IXED ion exchange/electrodialysis L liquid LDH layered double hydroxides LP permeability MCL maximum contaminant level MIL Matériaux de l′Institut Lavoisier MO Moringa oleifera MOF metal-organic framework MWCO molecular weight cut-off NDC 2,6-naphthalenedicarboxylate NF nanofiltration NFT nitrofurantoin NZF nitrofurazone PAC physical activated carbon Accepted Preprint Manuscript25 pHPZC point of zero charge PPCN polyethylene-glycol citrate-co-N-isopropylacrylamide PSAC palm shell activated carbon PSM post synthetic modifications PTA p-benzene-dicarboxylic acid RO reverse osmosis ROS reactive oxygen species rP pore radius RPDA photometric dispersion analysis S solid SAR sodium adsorption ratio SBU secondary building unit SDR sawdust raw SWHR seawater high rejection TAN total ammonia nitrogen TDS total dissolved solids TFC thin-film composite TIPA tris-4-imidazolyl-phenyl-amine TNP 2,4,6-trinitrophenol Accepted Preprint Manuscript26 ULPRO ultra-low-pressure reverse osmosis system WHO World Health Organisation WSR wheat straw raw XPS X-ray photoelectron spectroscopy XRD X-ray diffraction ZIF zeolitic imidazole framework ZnO zinc oxide ZrO2 zirconium oxide Chemical formulae Al(OH)3 aluminium hydroxide Al2(SO4)3 aluminium sulphate Al3+ aluminium cation AlnFm(OH)3n-m aluminium fluoride hydroxide complex Al-OH+ aluminol Als – O- hydroxylyzed aluminol, Als – OH neutral aluminol Als – OH2+ protonated aluminol As arsenic As(III) arsenite Accepted Preprint Mnuscript27 As(V) arsenate AsO33− arsenite AsO45− arsenate Ca2+ calcium cation Ca₅(PO₄)₃F fluorapatite Ca5(PO4)3OH hydroxyapatite CAC chemical activated carbon CaCO3 calcite CaF2 calcium fluoride/fluorspar CO2 carbon dioxide Co3O4 cobalt tetraoxide Cu copper F fluoride Fe3+ ferric/iron cation Fe3O4 iron oxide FeCl3 iron chloride H+ hydrogen ion H2AsO4- dihydrogen arsenate H3AsO3 arsenous acid Accepted Preprint Manuscript28 HAsO42- hydrogen arsenate Hg2+ mercuric ion Mg(H2gal) magnesium gallate Mg2+ magnesium cation MgF2 magnesium fluoride/sellaite MgO magnesium oxide Na3AlF6 sodium hexafluoroaluminate/cryolite Ni nickel NiO nickel oxide -OH hydroxyl group S sulphur Sn tin SnO2 tin oxide Symbols Co initial effluent concentration (mg/L or µg/L) Ct final effluent concentration (mg/L or µg/L) M adsorbent dosage (g) kTH Thomas model constant (mL/min.mg) qe adsorption capacity (mg/g) Q flow rate (mL/min) Accepted Preprint Manuscript29 T breakthrough time (min) Greek symbols Ψ constant surface electrical potential Conflict of Interest The authors declare that no conflict of interests exists in this review paper. Acknowledgements The review work is supported by research initiation grant offered to Dr. Somak Chatterjee, Department of Chemical Engineering, BITS Pilani, Pilani Campus and the research assistantship provided to Ms. Linisha Biswal by the BITS Pilani Institute. It is also a result of collaborative efforts between Dr. Somak Chatterjee, Dr. Joseph Goodwill and Dr. Christoph Janiak. Authors Responsibilities L. Biswal: Writing - Original Draft; Writing - Review & Editing, Joseph. E. Goodwill: Writing - Review & Editing, Christoph Janiak: Writing - Review & Editing, Somak Chatterjee: Conceptualization; Writing - Review & Editing; Supervision. Accepted Preprint Manuscript30 Reference [1] Gleick, P.H. Global Freshwater Resources: Soft-Path Solutions for the 21st Century. Science. 2003, 302, 1524-1528. https://doi.org/10.1126/science.1089967 [2] Edition, F. Guidelines for Drinking-water Quality. World Health, 2011. https://doi.org/10.1016/S1462-0758(00)00006-6 [3] The sustainable development goals report 2019. United Nations Publ. issued by Dep. Econ. Soc. Aff., 2019. [4] Singh, Jyoti, Priyanka Yadav, Ashok Kumar Pal, and Vishal Mishra. Water pollutants: Origin and status. In Sensors in Water Pollutants Monitoring: Role of Material, pp. 5-20. Springer, Singapore, 2020. https://doi.org/10.1007/978-981-15-0671-0_2 [5] Srivastav, Arun Lal. Chemical fertilizers and pesticides: role in groundwater contamination. In Agrochemicals detection, treatment and remediation, pp. 143-159. Butterworth-Heinemann, 2020. https://doi.org/10.1016/B978-0-08-103017-2.00006-4 [6] Berg, M., and J. E. Podgorski. Groundwater Assessment Platform (GAP): A new GIS tool for risk forecasting and mitigation of geogenic groundwater contamination. In Environmental Arsenic in a Changing World: Proceedings of the 7th International Congress and Exhibition on Arsenic in the Environment (AS 2018), July 1-6, 2018, Beijing, PR China, p. 5. CRC Press, 2019.https://doi.org/10.1201/9781351046633-2 [7] Alarcón-Herrera, M. T., Martin-Alarcon, D. A., Gutiérrez, M., Reynoso-Cuevas, L., Martín-Domínguez, A., Olmos-Márquez, M. A., Bundschuh, J. Co-occurrence, possible origin, and health-risk assessment of arsenic and fluoride in drinking water sources in Mexico: Geographical data visualization. Sci. Total Environ. 2020, 698, 134168. https://doi.org/10.1016/j.scitotenv.2019.134168 [8] Jha, S. K., and V. K. Mishra. Fluoride and arsenic in groundwater: occurrence and geochemical processes controlling Mobilisation. In Innovative saline agriculture, pp. 351-369. Springer, New Delhi, 2016. https://doi.org/10.1007/978-81-322-2770-0_16 [9] Mukherjee, A., Sengupta, M.K., Hossain, M.A., Ahamed, S., Das, B., Nayak, B., Lodh, D., Rahman, M.M., Chakraborti, D. Arsenic contamination in groundwater: a global perspective with emphasis on the Asian scenario. J. Health Popul. Nutr. 2006, 142-163. 10.3329/jhpn.v24i2.727 [10] Brindha, K., and Elango, L. Fluoride in groundwater: causes, implications and mitigation measures. Fluoride properties, applications and environmental management, 2011, 1, 111-136. https://www.novapublishers.com/catalog/product-info.php?products_id=15895 [11] Sandoval, M. A., Fuentes, R., Nava, J. L., Coreño, O., Li, Y., Hernández, J. H. Simultaneous removal of fluoride and arsenic from groundwater by electrocoagulation using a filter-press flow reactor with a three-cell stack. Sep. Purif. Technol., 2019, 208, 208-216. https://doi.org/10.1016/j.seppur.2018.02.018 Accepted Preprint Manuscript31 [12] Reardon, E. J., and Wang, Y. A limestone reactor for fluoride removal from wastewaters. Environ. Sci. Technol. 2000, 34, 3247-3253. https://doi.org/10.1021/es990542k [13] Zhou, C. C., Liu, C., Zhang, N. N., Cong, L. F., Pan, J. H., Peng, C. B. Fluorine in coal: the modes of occurrence and its removability by froth flotation. Int. J. Coal Prep. Util. 2018, 38, 149-161. https://doi.org/10.1080/19392699.2016.1215312 [14] Guo, Q., and Tian, J. Removal of fluoride and arsenate from aqueous solution by hydrocalumite via precipitation and anion exchange. Chem. Eng. J., 2018, 231, 121-131. https://doi.org/10.1016/j.cej.2013.07.025 [15] Richards, L. A., Richards, B. S., Rossiter, H. M. A., Schäfer, A. I. Impact of speciation on fluoride, arsenic and magnesium retention by nanofiltration/reverse osmosis in remote Australian communities. Desalination, 2009, 248, 177-183. https://doi.org/10.1016/j.desal.2008.05.054 [16] Salazar, H., Nunes-Pereira, J., Correia, D.M., Cardoso, V.F., Gonçalves, R., Martins, P.M., Ferdov, S., Martins, M.D., Botelho, G., Lanceros-Méndez, S. Poly (vinylidene fluoride-hexafluoropropylene)/bayerite composite membranes for efficient arsenic removal from water. Mater. Chem. Phys., 2016, 183, 430-438. https://doi.org/10.1016/j.matchemphys.2016.08.049 [17] Bhadja, V., Trivedi, J. S., Chatterjee, U. Efficacy of polyethylene interpolymer membranes for fluoride and arsenic ion removal during desalination of water via electrodialysis. RSC Adv., 2016, 6, 67118-67126. https://doi.org/10.1039/C6RA11450D [18] Li, W., Cao, C. Y., Wu, L. Y., Ge, M. F., Song, W. G. Superb fluoride and arsenic removal performance of highly ordered mesoporous aluminas. J. Hazard. Mater., 2011, 198, 143-150. https://doi.org/10.1016/j.jhazmat.2011.10.025 [19] Vázquez Mejía, G., Martínez-Miranda, V., Fall, C., Linares-Hernández, I., Solache- Ríos, M. Comparison of Fe–Al-modified natural materials by an electrochemical method and chemical precipitation for the adsorption of F− and As(V). Environ. Technol. Inno. 2016, 5, 558-568. https://doi.org/10.1080/09593330.2015.1074724 [20] Mohora, E., Rončević, S., Dalmacija, B., Agbaba, J., Watson, M., Karlović, E., Dalmacija, M. Removal of natural organic matter and arsenic from water by electrocoagulation/flotation continuous flow reactor. J. Hazard. Mater., 2012, 235, 257-264. https://doi.org/10.1016/j.jhazmat.2012.07.056 [21] Jamhour, R. M. A. Q. New inorganic ion-exchange material for the selective removal of fluoride from potable water using ion-selective electrode. Am. J. Environ. Sci., 2005, 1, 1-4. https://doi.10.3844/ajessp.2005.1.4 [22] Kim, J., and Benjamin, M. M. Modeling a novel ion exchange process for arsenic and nitrate removal. Water Res., 2004, 38, 2053-2062. https://doi.org/10.1016/j.watres.2004.01.012 [23] Durmaz, F., Kara, H., Cengeloglu, Y., Ersoz, M. Fluoride removal by Donnan dialysis with anion exchange membranes. Desalination, 2005, 177, 51-57. https://doi.org/10.1016/j.desal.2004.11.016 Accepted Preprint Manuscript32 [24] Tian, Y., Wu, M., Liu, R., Wang, D., Lin, X., Liu, W., Ma, L., Li, Y., Huang, Y. Modified native cellulose fibers—A novel efficient adsorbent for both fluoride and arsenic. J. Hazard. Mater., 2011, 185, 93-100. https://doi.org/10.1016/j.jhazmat.2010.09.001 [25] Hasan, Z., and Jhung, S. H. Removal of hazardous organics from water using metal-organic frameworks (MOFs): plausible mechanisms for selective adsorptions. J. Hazard. Mater., 2015, 283, 329-339. https://doi.org/10.1016/j.jhazmat.2014.09.046 [26] Yaghi, O. M., Li, G., Li, H. Selective binding and removal of guests in a microporous metal–organic framework. Nature, 1995, 378, 703-706. https://doi.org/10.1038/378703a0 [27] Bobbitt, N. S., and Snurr, R. Q. Molecular modelling and machine learning for high-throughput screening of metal-organic frameworks for hydrogen storage. Mol. Simulat. 2019, 45, 1069-1081. https://doi.org/10.1080/08927022.2019.1597271 [28] Li, T., Sullivan, J. E., Rosi, N. L. Design and preparation of a core–shell metal–organic framework for selective CO2 capture. J. Am. Chem. Soc. 2013, 135, 9984-9987. https://doi.org/10.1021/ja403008j [29] Bobbitt, N. S., Mendonca, M. L., Howarth, A. J., Islamoglu, T., Hupp, J. T., Farha, O. K., Snurr, R. Q. Metal–organic frameworks for the removal of toxic industrial chemicals and chemical warfare agents. Chem. Soc. Rev. 2017, 46, 3357-3385. https://doi.org/10.1039/C7CS00108H [30] Jayaramulu, K., Narayanan, R. P., George, S. J., Maji, T. K. Luminescent microporous metal–organic framework with functional lewis basic sites on the pore surface: specific sensing and removal of metal ions. Inorg. Chem. 2012, 51, 10089-10091. https://doi.org/10.1021/ic3017547 [31] Haque, E., Jun, J. W., Jhung, S. H. Adsorptive removal of methyl orange and methylene blue from aqueous solution with a metal-organic framework material, iron terephthalate (MOF-235). J. Hazard. Mater., 2011, 185, 507-511. https://doi.org/10.1016/j.jhazmat.2010.09.035 [32] Jung, B. K., Hasan, Z., Jhung, S. H. Adsorptive removal of 2, 4-dichlorophenoxyacetic acid (2, 4-D) from water with a metal–organic framework. Chem. Eng. J. 2013, 234, 99-105. https://doi.org/10.1016/j.cej.2013.08.110 [33] Lin, K. Y. A., and Chang, H. A. Efficient adsorptive removal of humic acid from water using zeolitic imidazole framework-8 (ZIF-8). Water Air Soil Poll. 2015, 226, 10. https://doi.org/10.1007/s11270-014-2280-7 [34] Li, J., Wang, X., Zhao, G., Chen, C., Chai, Z., Alsaedi, A., Hayat, T., Wang, X. Metal–organic framework-based materials: superior adsorbents for the capture of toxic and radioactive metal ions. Chem. Soc. Rev. 2018, 47, 2322-2356. https://doi.org/10.1039/C7CS00543A [35] Zhang, N., Yang, X., Yu, X., Jia, Y., Wang, J., Kong, L., Jin, Z., Sun, B., Luo, T., Liu, J. Al-1, 3, 5-benzenetricarboxylic metal–organic frameworks: A promising adsorbent for defluoridation of water with pH insensitivity and low aluminum residual. Chem. Eng. J. 2014, 252, 220-229. https://doi.org/10.1016/j.cej.2014.04.090 Accepted Preprint Manuscript33 [36] Zhu, B.J., Yu, X.Y., Jia, Y., Peng, F.M., Sun, B., Zhang, M.Y., Luo, T., Liu, J.H., Huang, X.J. Iron and 1, 3, 5-benzenetricarboxylic metal–organic coordination polymers prepared by solvothermal method and their application in efficient As (V) removal from aqueous solutions. J. Phys. Chem. C 2012, 116, 8601-8607. https://doi.org/10.1021/jp212514a [37] World Health Organization. Progress on household drinking water, sanitation and hygiene 2000-2017: special focus on inequalities. World Health Organization, 2019. [38] Chouhan, S., and Flora, S. J. Arsenic and fluoride: two major ground water pollutants. Indian J. Exp. Biol. 2010, 48, 666. http://nopr.niscair.res.in/handle/123456789/9735 [39] Edmunds, W. Mike, and Pauline L. Smedley. \"Fluoride in natural waters.\" In Essentials of medical geology, pp. 311-336. Springer, Dordrecht, 2013. https://doi.org/10.1007/978-94-007-4375-5_13 [40] Agarwal, M., Rai, K., Shrivastav, R., Dass, S. Fluoride speciation in aqueous suspensions of montmorillonite and kaolinite. Toxicol. Environ. Chem. 2002, 82, 11-21. https://doi.org/10.1080/713746660 [41] Aoba, T., and Fejerskov, O. Dental fluorosis: chemistry and biology. Crit. Rev. Oral Biol. Med. 2002, 13, 155-170. https://doi.org/10.1177%2F154411130201300206 [42] Srivastava, S., and Flora, S. J. S. Fluoride in Drinking Water and Skeletal Fluorosis: A Review of the Global Impact. Curr. Environ. Health Rep. 2020, 1-7. https://doi.org/10.1007/s40572-020-00270-9 [43] Nayak, B., Roy, M. M., Das, B., Pal, A., Sengupta, M. K., Prasad De, S., Chakraborti, D. Health effects of groundwater fluoride contamination. Clin. Toxicol. (Phila.), 2009, 47, 292-295. https://doi.org/10.1080/15563650802660349 [44] Ali, S., Thakur, S. K., Sarkar, A., Shekhar, S.Worldwide contamination of water by fluoride. Environ. Chem. Lett. 2016, 14, 291-315. https://doi.org/10.1007/s10311-016-0563-5 [45] Senior, Lisa A., and Ronald A. Sloto. Arsenic, boron, and fluoride concentrations in ground water in and near diabase intrusions, Newark Basin, Southeastern Pennsylvania. No. 2006-5261. US Geological Survey, 2006. https://doi.org/10.3133/sir20065261 [46] Patel, K. S., Shrivas, K., Brandt, R., Jakubowski, N., Corns, W., Hoffmann, P. Arsenic contamination in water, soil, sediment and rice of central India. Environ. Geochem. Health. 2005, 27, 131-145. https://doi.org/10.1007/s10653-005-0120-9 [47] Garelick, H., Jones H., Dybowska A., Valsami-Jones. E. Arsenic pollution sources. Reviews of Environmental Contamination Volume 197 (2009): 17-60.https://doi.org/10.1007/978-0-387-79284-2_2 [48] Ng, J. C., Wang, J., and Shraim, A. A global health problem caused by arsenic from natural sources. Chemosphere, 2003, 52, 1353-1359. 10.1016/S0045-6535(03)00470-3. Accepted Preprint Manuscript34 [49] Huq, S. I., Joardar, J. C., Parvin, S., Correll, R., Naidu, R. Arsenic contamination in food-chain: transfer of arsenic into food materials through groundwater irrigation. J. Health Popul. Nutr. 2006, 24, 305. 10.3329/jhpn.v24i3.716 [50] Rahman, M. M., Ng, J. C., Naidu, R. Chronic exposure of arsenic via drinking water and its adverse health impacts on humans. Environ. Geochem. Health. 2009, 31, 189-200. https://doi.org/10.1080/09603120220129346 [51] Organization, W.H. (2010). Chemical fact sheets. Guidel. Drink.Organization [52] Wang, L. K., Hung, Y. T., Shammas, N, K., eds. Physicochemical treatment processes. Vol. 3. Totowa, NJ: Humana Press, 2005.https://doi.org/10.1385/1-59259-820-x:141 [53] Turner, B. D., Binning, P. Stipp, S. L. S. Fluoride removal by calcite: evidence for fluorite precipitation and surface adsorption. Environ. Sci. Technol. 2005, 39, 9561-9568. https://doi.org/10.1021/es0505090 [54] Ruiping, L., Xing, L., Shengji, X., Yanling, Y., Rongcheng, W., Guibai, L. Calcium-Enhanced Ferric Hydroxide Co-Precipitation of Arsenic in the Presence of Silicate. Water Environ. Res. 2007, 79, 2260-2264. https://doi.org/10.2175/106143007X199324 [55] Dungkaew, W., Haller, K.J., Flood, A.E., Scamehorn, J.F. Arsenic removal by precipitation with Calcium phosphate hydroxyapatite. In Advanced Materials Research, vol. 506, pp. 413-416. Trans Tech Publications Ltd, 2012. https://doi.org/10.4028/www.scientific.net/AMR.506.413 [56] Emamjomeh, M. M., and Sivakumar, M. Fluoride removal by a continuous flow electrocoagulation reactor. J. Environ. Manage. 2009, 90, 1204-1212. https://doi.org/10.1016/j.jenvman.2008.06.001 [57] Khatibikamal, V., Torabian, A., Janpoor, F., Hoshyaripour, G. Fluoride removal from industrial wastewater using electrocoagulation and its adsorption kinetics. J. Hazard. Mater. 2010, 179, 276-280. https://doi.org/10.1016/j.jhazmat.2010.02.089 [58] Jafari, A., Mahvi, A. H., Godini, H., Rezaee, R., Hosseini, S. S. Process optimization for fluoride removal from water by Moringa oleifera seed extract. Fluoride. 2014, 47, 152-60. [59] Song, S., Lopez-Valdivieso, A., Hernandez-Campos, D. J., Peng, C., Monroy-Fernandez, M. G., Razo-Soto, I. Arsenic removal from high-arsenic water by enhanced coagulation with ferric ions and coarse calcite. Water Res. 2006, 40, 364-372. https://doi.org/10.1016/j.watres.2005.09.046 [60] Balasubramanian, N., and Madhavan, K. Arsenic removal from industrial effluent through electrocoagulation. Chem. Eng. Technol. 2001, 24, 519-521. https://doi.org/10.1002/1521-4125(200105)24:5<519::AID-CEAT519>3.0.CO;2-P [61] Vasudevan, S., Lakshmi, J., Sozhan, G. Studies on the removal of arsenate by electrochemical coagulation using aluminum alloy anode. CLEAN–Soil, Air, Water. 2010, 38, 506-515. https://doi.org/10.1002/clen.201000001 Accepted Preprint Manuscript35 [62] Ndiaye, P. I., Moulin, P., Dominguez, L., Millet, J. C., Charbit, F. Removal of fluoride from electronic industrial effluent by RO membrane separation. Desalination, 2005, 173, 25-32. https://doi.org/10.1016/j.desal.2004.07.042 [63] Liu, X., Hu, Z., Zhu, C., Wen, G., Meng, X. Removal of fluoride and total dissolved solids from coalbed methane produced water with a movable ultra-low-pressure reverse osmosis system. Desalination Water Treat. 2013, 51, 4359-4367. https://doi.org/10.1080/19443994.2012.742853 [64] Akin, I., Arslan, G., Tor, A., Cengeloglu, Y., Ersoz, M. Removal of arsenate [As (V) and arsenite [As (III) from water by SWHR and BW-30 reverse osmosis. Desalination. 2011, 281, 88-92. https://doi.org/10.1016/j.desal.2011.07.062 [65] Hu, K., and Dickson, J. M. Nanofiltration membrane performance on fluoride removal from water. J. Membrane Sci. 2006, 279, 529-538. https://doi.org/10.1016/j.memsci.2005.12.047 [66] Chakrabortty, S., Roy, M., Pal, P. Removal of fluoride from contaminated groundwater by cross flow nanofiltration: transport modeling and economic evaluation. Desalination. 2013, 313, 115-124. https://doi.org/10.1016/j.desal.2012.12.021 [67] Nguyen, C. M., Bang, S., Cho, J., Kim, K. W. Performance and mechanism of arsenic removal from water by a nanofiltration membrane. Desalination. 2009, 245, 82-94. 10.1016/j.desal.2008.04.047 [68] Harisha, R. S., Hosamani, K. M., Keri, R. S., Nataraj, S. K., Aminabhavi, T. M. Arsenic removal from drinking water using thin film composite nanofiltration membrane. Desalination. 2010, 252, 75-80. 10.1016/j.desal.2009.10.022 [69] Kabay, N., Arar, Ö., Samatya, S., Yüksel, Ü., Yüksel, M. Separation of fluoride from aqueous solution by electrodialysis: Effect of process parameters and other ionic species. J. Hazard. Mater. 2008, 153, 107-113. https://doi.org/10.1016/j.jhazmat.2007.08.024 [70] Ergun, E., Tor, A., Cengeloglu, Y., Kocak, I. Electrodialytic removal of fluoride from water: Effects of process parameters and accompanying anions. Sep."},"Similar Full Paper":{"kind":"string","value":"{'MOF-derived Carbon-Based Materials for Energy-Related Applications.': \"Title: MOF-derived Carbon-Based Materials for Energy-Related Applications.\\\\nYap, Min Hui and Fow, Kam Loon and Chen, George Zheng (2017) Synthesis and applications of MOF - derived porous nanostructures. Green Energy & Environment, 2 (3). pp. 218-245. ISSN 2468-0257 Access from the University of Nottingham repository: http://eprints.nottingham.ac.uk/43420/8/1-s2.0-S2468025717300638-main.pdfCopyright and reuse: The Nottingham ePrints service makes this work by researchers of the University of Nottingham available open access under the following conditions.This article is made available under the Creative Commons Attribution Non-commercial No Derivatives licence and may be reused according to the conditions of the licence. For more details see: http://creativecommons.org/licenses/by-nc-nd/2.5/A note on versions: The version presented here may differ from the published version or from the version of record. If you wish to cite this item you are advised to consult the publisher’s version. Please see the repository url above for details on accessing the published version and note that access may require a subscription.For more information, please contact eprints@nottingham.ac.ukAvailable online at www.sciencedirect.comScienceDirectGreen Energy & Environment 2 (2017) 218e245www.keaipublishing.com/geeReview articleSynthesis and applications of MOF-derived porous nanostructuresMin Hui Yap a, Kam Loon Fow a,*, George Zheng Chen a,b,*a Department of Chemical and Environmental Engineering, and Energy Engineering Research Group, Faculty of Science and Engineering, University ofNottingham Ningbo China, 199 Taikang East Road, Ningbo 315100, Chinab Department of Chemical and Environmental Engineering, and Advanced Materials Research Group, Faculty of Engineering, University of Nottingham,Nottingham NG7 2RD, UKReceived 31 March 2017; revised 17 May 2017; accepted 17 May 2017Available online 25 May 2017AbstractMetal organic frameworks (MOFs) represent a class of porous material which is formed by strong bonds between metal ions and organiclinkers. By careful selection of constituents, MOFs can exhibit very high surface area, large pore volume, and excellent chemical stability.Research on synthesis, structures and properties of various MOFs has shown that they are promising materials for many applications, such asenergy storage, gas storage, heterogeneous catalysis and sensing. Apart from direct use, MOFs have also been used as support substrates fornanomaterials or as sacrificial templates/precursors for preparation of various functional nanostructures. In this review, we aim to present themost recent development of MOFs as precursors for the preparation of various nanostructures and their potential applications in energy-relateddevices and processes. Specifically, this present survey intends to push the boundaries and covers the literatures from the year 2013 to early 2017,on supercapacitors, lithium ion batteries, electrocatalysts, photocatalyst, gas sensing, water treatment, solar cells, and carbon dioxide capture.Finally, an outlook in terms of future challenges and potential prospects towards industrial applications are also discussed.© 2017, Institute of Process Engineering, Chinese Academy of Sciences. Publishing services by Elsevier B.V. on behalf of KeAi Communi-cations Co., Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).Keywords: Metal organic frameworks; Porous nanostructures; Supercapacitors; Lithium ion batteries; Heterogeneous catalyst1. Introduction1.1. Metal organic frameworks (MOFs)Porous materials such as porous ceramics, zeolites, acti-vated charcoals, porous metal, polymer foams, and porousglass are being used in many ways in our daily lives. Due totheir well-known properties and wide range applications, thefield of porous materials, in particular the nanoporous mate-rials, has undergone rapid development in the past two* Corresponding authors. Department of Chemical and Environmental En-gineering, and Advanced Materials Research Group, Faculty of Engineering,University of Nottingham, Nottingham NG7 2RD, UK.E-mail addresses: Kam-Loon.Fow@nottingham.edu.cn (K.L. Fow),george.chen@nottingham.ac.uk (G.Z. Chen).http://dx.doi.org/10.1016/j.gee.2017.05.0032468-0257/© 2017, Institute of Process Engineering, Chinese Academy of SciencesLtd. This is an open access article under the CC BY-NC-ND license (http://creativdecades. Among the recent developed porous materials, metalorganic frameworks (MOFs) are distinct from other traditionalporous materials because of their high porosity and thermalstability. Formed by the three-dimensional crystalline assem-bly of inorganic metal ions and organic ligands, MOFs enableflexible structure design of which well-defined pore sizes,surfaces areas and functionalities can be tailored by selectingdifferent building blocks. This high degree of customizabilityof MOFs properties has attracted the interest of many re-searchers. To date, there are more than 20,000 differentstructures of MOFs being reported and studied [1]. A fewexamples of different MOFs structures are illustrated in Fig. 1.Depending on the final structures and properties, MOFs maybe prepared using several distinct synthetic methods such as:slow diffusion [3], hydrothermal (solvothermal) [4], electro-chemical [5], mechanochemical [6], microwave assisted heat-ing and ultrasound [7]. These synthesis methods and formation. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co.,ecommons.org/licenses/by-nc-nd/4.0/).List of abbreviationsMOFs Metal organic frameworksZIF Zeolitic imidazolate frameworkEDLCs Electric double-layer capacitorsLIBs Lithium ion batteriesMWCNTs Multi-walled carbon nanotubesTEM Transmission electron microscopyNPC Nitrogen-doped porous carbonNPM Non-precious metalAQ AnthraquinoneNQ 1,4-napthoquinoneTCBQ TetrachlorobenzoquinoneBDC 1,4-benzenedicarboxylic acidBTC 1,3,5-benzenetricarboxylic acidNTCDA 1,4,5,8-naphthalenetetracarboxylicdianhydridePTCDA Perylene-3,4,9,10-tetracarboxylicdianhydrideMMT MontmorilloniteRHE Reversible hydrogen electrodeSCE Saturated calomel electrode219M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245mechanisms of MOFs have been comprehensively reviewed bySeoane and co-workers recently [8]. A wide range of potentialapplications of MOFs, ranging from gas (e.g. hydrogen) stor-age and separation, sensing, catalysis, to drug delivery, has alsobeen reviewed [9,10]. After nearly three decades since the firstreport on synthesis MOFs [11], a few MOFs are now producedcommercially. One of the most prominent commercializedMOFs is Cu-BTC (also known as HKUST-1) produced byBASF (marketed under the trademark BASOLITE® C 300) andsold by Sigma Aldrich [2]. Among other companies which holdpatents for large scale synthesis of MOFs are MOF Technol-ogies, Ford Global Technologies, Toyota, and Strem Chem-icals. A search on the World Intellectual Property Organization(WIPO) database revealed a constant increment of the numberof patents published by the world-wide research community.This indicates that MOFs are gaining a considerable mo-mentum towards commercial applications. The number ofpatent publications from the years 2007e2017 and the numberof patents filed by each country are shown in Figs. 2 and 3,respectively. These patents cover the production and applica-tions of various MOFs or their composites. Fig. 3 also high-lights that the top two highest numbers of patents are filed byinnovators from United States and China. This is a clear evi-dence that the researchers from these two countries arecommitted to the exploration of the commercial opportunitiesof MOFs. On the other hand, according to the WIPO database,the company which currently owns the highest number ofpatents is BASF, with 133 patents published so far.In addition, the huge library of MOFs structures and theoptimized synthesis methods are helpful for researchers toexplore some other potential applications of MOFs, such asthe usage of MOFs as sacrificial materials for the synthesis ofvarious nanostructures. This has opened a new direction inapplication of MOFs, and might contribute to a better under-standing of the properties of porous materials. It is evident thatmany of the papers published from the years 2008e2013 areshort communications or research work focusing on the syn-thesis of various new nanostructures from MOFs and deter-mination of the basic physical properties of the resultantstructures. Very few have measured the performance of thesematerials against specific applications and study their recy-clability. The synthesis method of nanostructures from MOFshas been reviewed thoroughly by Mai et al. [12]. However, tothe best of our knowledge, there is not yet a review that isspecifically devoted to the synthesis and characterizations ofMOFs-derived nanostructures for applications in energyrelated materials, devices, and processes. For this very reason,we intend to seize this opportunity to review on recent ad-vances in MOF-derived nanostructures for energy-related ap-plications from the years 2013e2017, and to anticipate theprospects of these MOF-derived nanostructures.1.2. Synthesis of various nanostructures from MOFsPorous carbon, metals, metal oxides, and their multicom-ponent hybrids are important inorganic materials for energyand environmental applications. Dependent on the desiredapplication purposes, porous materials can be prepared byseveral synthetic approaches such as hard templating and softtemplating. Hard templating method involves the frameworkprecursors filling the cavities present in the structured solidtemplate of which the template can be removed from theporous structure after synthesis. In comparison, soft templatingmethod involves a more subtle physical or chemical in-teractions between framework source and template whichdirect the self-assembly synthesis to allow better control of thematerial properties [13]. When compared between these twotemplating approaches, soft templating provides a more suc-cessful pathway for the synthesis of ordered and disorderedporous matrices [14]. In the soft templating route, porousmaterials are generally synthesized using the solvothermalmethod. This solvothermal method is also the most preferredmethod for large scale production, owing to its relativesimplicity and scalability. Aside being used for the synthesis ofzeolite, the concept of solvothermal synthesis method has alsobeen applied with great success for the synthesis of MOFs.In the context of synthesizing various nanostructures fromMOFs, MOFs can act as a precursor in which the metalcomponents provide an intrinsic metal source to derivenanostructures of metals or metals oxides, as well as a self-sacrificing template, in which the organic components canbe used as a carbon source to prepare nanoporous carbon [15].In general, certain MOF structures, such as Cu-BTC andMOF-5, are collapsed during the carbonization process,whereas other MOF structures, such as ZIF-8 and ZIF-67, arebetter in providing a template to guide the formation of poresby allowing evaporation of confined organic moistures duringpyrolysis, resulting in a spongy-pore system. Nevertheless, ahomogeneous distribution of nanoparticles in their respectiveMOF-5 HKUST-1 [CuSiF6(4,4’-bpy)] MOP-1MOF-177MIL-88MIL-53MIL-47ELM-11Cr-MIL-100 Cr-MIL-101 Ni-CPO-27 UiO-66 ZIF-8NOTT-116UMCM-2[Be12(OH)12(BTB)4]DO-MOFPCN-14MOF-200 UTSA-20 IRMOF-74-XI NU-125MOF-14Fig. 1. Illustration of nanoporous structures of different MOFs synthesized by different research groups. Reproduced from Ref. [2], with permission from the RoyalSociety of Chemistry.220 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245medium is expected. Some of the advantages of using MOFsas a precursor/sacrificial template for preparation of variousnanostructures include, but not limited to,i) the fabrication of MOF-derived nanostructures withdesired topological textures and material properties aremade possible with the controlled-calcination of MOFsii) undesired structural collapse of the framework duringcalcination can be reduced with the use of MOFs as tem-plates since they generally show structural robustness [16]iii) allow easy functionalization with other heteroatoms andmetal/metal oxides, therefore able to increase overallperformance and efficiencyiv) the preparation of MOFs can be carried out under mildconditions and simple processes274758 575073 7510015287170204060801001201401602007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017Number of Patent PublicaƟonsYearFig. 2. Number of patent publications from 2007 to 2017, according to the World Intellectual Property Organization database. (Search Terms: metal organicframework).19819017687634010 7 3 2 2 1 1050100150200250United StatesChina Patent Cooperation TreatyEuropean Patent OfficeRepublic of KoreaCanada Mexico Germany Singapore United KingdomJapan Isreal PortugalNumber of Patent PublicationsCountriesFig. 3. Number of patent publications by countries according to the World Intellectual Property Organization database, from 2007 to 2017. (Search Terms: metalorganic framework).221M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245Benefiting from these advantages, MOF-derived nano-structures have narrower distributions of pore volumes, andhigher specific surface area, and have more variations inmorphology, such as nanocages, hollow spheres, and hollowpolyhedrons, compared to other nanostructures. Additionally,this synthesis approach also helps solve the potential in-compatibility issue between electroactive material and 3Dsubstrate support through some unique growth and decompo-sition mechanisms. However, on-going research is aiming toaddress and overcome the limitations of using MOF as pre-cursors/templates. Some of these limitations include poorcontrol of the pore size of the MOF-derived nanostructuresdue to the lack of understanding of the MOF decompositionmechanisms, and relatively high calcination temperatures arerequired.Notably, by subjecting MOF precursors of various struc-tures and crystallinity to different calcination temperatures,Calcination in air Pyrolysis Metal Organic Frameworks Fig. 4. Overview of different types of nanosmaterial with different topologies, crystallographic phases,and porosities can be obtained. Fig. 4 shows the overview ofdifferent types of nanostructures synthesized from thermaltreatment of MOF precursors. It is worth mentioned that ion-exchange method has also emerged recently as an alternativemethod to thermal treatment for preparation of various nano-structures from MOF precursors. Although more chemicalsand experimental steps were involved in the ion-exchangemethod, nevertheless, the as-synthesized nanostructures fromthis method have demonstrated some exciting properties[30,32,58].Overall, this article aims to present a comprehensive reviewon the recent development in the synthesis and applications ofvarious nanostructures derived from different MOFs. In thesubsequent sections, we will summarize some recent excitingexamples of these applications across the field of electro-chemical energy storage, catalysis, sensing, andwater treatment.Multi-metal oxides, metal/carbon, Metal/Metal Oxide Metal oxide/carbon, Metal-Metal Oxide/Carbon, Metal carbide, Metal sulfide tructures derived from MOFs Precursor.222 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e2452. Applications of MOF-derived nanostructures inelectrochemical energy storageThe demand for efficient use of clean and renewable en-ergy, as well as, the popularization of portable electronics andelectric vehicles has prompted the development of energystorage materials, especially for electrochemical energy stor-age [17]. The use of nanomaterials for electrochemical energystorage has emerged as a promising approach [18]. Given theirunique electrical, mechanical properties, and high surfacearea, nano-sized active materials are expected to bridge thegap towards the realization of the next generation of energystorage device [19]. For being competitive, these materialsalso need to have long cycle-life, good charge/discharge effi-ciency, good retention of charge and adequate operate voltageto ensure high energy and power density of the storage devices[20]. Many investigations had been carried out in the searchfor a better synthesis method for producing materials to caterdifferent electrochemical energy storage purposes. The directpyrolysis/carbonization of MOFs is a new method for syn-thesis of porous materials for energy storage. The use ofpristine MOF and MOF-derived structures for electrochemicalenergy storage and conversion has been reviewed by Xia andco-workers in early 2015 [17]. Nonetheless, this review willcover the MOF-derived nanostructures and their applicationsas electrode materials for supercapacitors and LIBs, with afocus of more recent works published between 2015 and 2017.In order to compare and show the classifications of variouselectrochemical energy storage devices, a graphical illustra-tion of a simplified Ragone plot of specific power versusspecific energy is shown in Fig. 5.Specific energy (Wh/kg)10 102 103InternalcombustionengineRechargeablebatteryRedox flow battery& Fuel cellSupercapatteryEDL supercapacitor10102103Specific power (W/kg)Ragone PlotFig. 5. Illustrative Ragone plots of specific power versus specific energy for thevarious electrochemical energy storage devices and internal combustion en-gine [21].2.1. SupercapacitorsA capacitor is a passive component designed to store en-ergy in an electric field. It generally consists of two conductingplates separated by an insulator (dielectric) [21,22]. A super-capacitor (electrochemical capacitor), on the other hand, is anideal electrical energy storage device with higher powercapability, longer cyclic life, and higher packaging flexibilitycompared to traditional energy storage devices [23]. Presently,there are two types of supercapacitors, namely: the electricdouble-layer capacitors (EDLCs) and pseudocapacitors (whichemploy a reversible Faradaic reaction to store charges)[21,24]. Carbon materials are commonly used as electrodematerials for EDLCs. However, they do not have very highspecific capacitance despite their high surface area. Theincorporation of heteroatoms such as nitrogen or oxygen intothe carbon structures can improve the capacitive storage of thecarbonaceous materials. This is well demonstrated by Zhongand co-workers, who prepared a series of nitrogen-dopedporous carbon (NPC) from ZIF-8 (prepared from zinc saltsand 2-methylimidazole) with the addition of carbon sources(melamine, urea, xylitol and sucrose) [25]. Their resultsproved that the introduction of additional carbon sourceswould form a protective layer around the samples, thusreducing their nitrogen loss during carbonization. This wasevident from the fact that the NPC sample with the highestnitrogen content (4.5 wt.%) had the best capacitive behavior(285.8 F g\\\\x011 at a mass normalized current of 0.1 A g\\\\x011).Aside from nitrogen doping and carbon capping, anotherstrategy that could increase the performance of MOF-basedcarbon materials is support modification and redox-active mol-ecules functionalization. In the year 2016, Wang's team coatedZIF-8 derived carbon onto multi-walled carbon nanotubes(MWCNTs) to form necklace-like structures [26]. The sche-matic illustration for the synthesis of C@MWCNTs, its resultantTEM image, and the specific capacitances at different currentdensities are illustrated in Fig. 6. The synthesis began by stirringa mixture solution of ZIF-8 and MWCNTs, kept still for 24 h,then followed by annealing at 800 \\\\x03C for 3 h. After carboniza-tion, the porous carbons were embedded on the surface ofcrystalline MWCNTs, as evident from the TEM image (Fig. 6).When compared to other coreeshell structures and MWCNTsbased coreeshell structures, this MOF-derived C@MWCNTsprovide some extra accessible surfaces for ion transports whichdoes not agglomerate or restack. The MOF-derived nano-structures have some improved performances, such as highspecific capacitance (326 F g\\\\x011 at a mass normalized current of1 A g\\\\x011), good rate capability, and excellent cycling stability(99.7% capacitance retention after 10,000 cycles).Guo's group had attempted the functionalization of someNPC materials by the incorporation of anthraquinone (AQ),1,4-napthoquinone (NQ), and tetrachlorobenzoquinone(TCBQ) to the NPC framework via a noncovalent interactionsmethod [27]. Results indicated that AQ-NPC had the bestcurrent response in the negative potential region with a spe-cific capacitance of 373 F g\\\\x011 at a mass normalized current of1 A g\\\\x011, whereas a variant combination, (TCBQ þ NQ)-NPC,(a)(c)surface modification+PVPoriented carbonizationC-ZIF-8@MWCNT necklaceC-ZIF-8/MWCNT mixtureMWCNT80.6%68.9%62.9%62.5%C-ZIF-8growthScan rate (mV s-1)04080120160200240280320(b).Specific capacitance (F g-1).0 40 8020 60 100Fig. 6. Schematic illustration for (a) the synthesis of C@MWCNTs, (b) its resultant TEM image, and (c) the specific capacitances at different current densities.Reproduced from Ref. [26], with permission from Elsevier.223M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245showed a current response in the positive potential region.Together they generated a potential self-matching behavior inthe two-electrode configuration, which led to a high perfor-mance asymmetrical supercapacitor with a high specific en-ergy value of 23.5 Wh kg\\\\x011. More importantly, thisfunctionalized method had enhanced the overall capacitanceof carbon by 1.4 times.In general, the performance of carbon-based materials asEDLC is affected by pore size distribution and pore volume.On the other hand, the redox active species for Faradaic ca-pacitors are dependent on the size of the redox active species,porosity, crystallinity and surface area of the electrode mate-rials for energy storage [17]. Redox active species are transi-tion metal oxides or conductive polymers which have a storagemechanism based on reversible Faradaic redox reactions. Thepreparation of redox active species using MOF precursors areless common compared to the preparation of carbon materialsfrom MOFs.In 2014, Zhang's team described the calcination of ZIF-67(prepared from cobalt salts and 2-methylimidazole) in air at450 \\\\x03C for 30 min (with a slow heating rate of 1 \\\\x03C/min) forthe preparation of porous hollow Co3O4 with rhombicdodecahedral structures [28]. This hollow rhombic structureprovided structural stability for the charging/discharging cyclewhereas the highly mesoporous Co3O4 structures enabled thefacile penetration of electrolyte and guaranteed a relativelyhigh electroactive surface area. These properties had suc-cessfully increased the overall rate capacity and cycle life ofthe hollow Co3O4 nanoparticles. These nanoparticles reporteda remarkable specific capacitance of 1100 F g\\\\x011 at a high massnormalized current of 12.5 A g\\\\x011.The incorporation of Ni2þ cations, S2\\\\x01 anions or OH\\\\x01 an-ions could increase electrical conductivity and aid the forma-tion of new active sites. In 2015, Hu's group synthesizedCo3O4/NiCo2O4 double shelled nanocages from ZIF-67 andNieCo layered double hydroxides (source of Ni2þ cations)[29]. The formation process of these nanocages is shown inFig. 7. The as-synthesized materials were evaluated as elec-trodes for pseudo capacitors and electrocatalysts for oxygengeneration. Results showed that the Co3O4/NiCo2O4 nanocagesexhibited a high capacitance of 972 F g\\\\x011 at 5 A g\\\\x011 and evenan impressive 615 F g\\\\x011 of capacitance at 50 A g\\\\x011. A superiorcycling stability of 12,000 cycle life was also achieved. Besidesthat, as an electrocatalyst, the bimetallic nanocages delivered acurrent density of 10 mA cm\\\\x012 at a potential of 1.57 V vs.reversible hydrogen electrode (RHE). The greatly enhancedperformance was attributed to the composition and complexstructure of the double shelled nanocages and the incorporationof Ni2þ cations. On the other hand, the incorporation of S2\\\\x01anions to the nanostructures was demonstrated by Jiang's teamwho fabricated hollow CoS nanocages from ZIF-67 templateand thioacetamide using ion-exchange method [30]. Eventhough the cycling stability was not high, however, these CoSpolyhedral nanocages had the highest reported specificcapacitance among those nanostructures discussed in thisSection 2.1 (~1400 F g\\\\x011 at a mass normalized current of10 A g\\\\x011). The study had successfully proved the importanceof S2\\\\x01 hydrolyzing of ZIF-67 and the effectiveness of usingMOF template in producing nanostructures with high per-forming pseudo capacitance properties. Hu and co-workersworked on a similar method of using ZIF-67 to produce CoSnanocages but added an additional step of reaction with waterIZIF-67 ZIF-67/Ni-Co LDHyolk-shelled structuresCo3O4/NiCo2O4DSNCsIIFig. 7. Schematic illustration of the formation process of Co3O4/NiCo2O4 double shelled nanocages. Reproduced from Ref. [29], with permission from AmericanChemical Society.224 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245(OH\\\\x01 anions) to obtain double-shelled hollow CoS nano-structures surrounded by outer CoS nanosheet-constructedshells [31]. Without the presence of carbon, the double-shelled hollow structures had showed a large specific surfacearea of 11 m2 g\\\\x011. Not only did the double-shelled assemblysupply more active sites for electrochemical reactions, it alsoresulted in good structural robustness to enhance electro-chemical stability. The stability of CoS was demonstrated whena specific capacitance of 765.6 F g\\\\x011 (at a mass normalizedcurrent of 10 A g\\\\x011) was achieved after 10,000 cycles.Recently, Guan's team carried out two ion-exchange reactions(with S2\\\\x01 anions and then with Ni2þ cations) to obtain onion-like NiCo2S4 particle from a Co-MOF [32]. Remarkably, theseven-layered onion-like structure delivered better rate per-formance, higher energy density, higher power density, andcomparable capacitance retention (after tested for 10,000 cy-cles) than those NieCo sulfide based electrodes derived fromnon-MOF template reported in Ref [33e35]. Benefiting fromthe complex composition and multi-shelled structure, a specificcapacitance of 765.6 F g\\\\x011 was achieved after 10,000 cycles ata mass normalized current of 10 A g\\\\x011.Furthermore, among the novel MOF-derived redoxactive species, the binary mixed metal/metal oxide@carboncomposites prepared by Wang and co-workers had showedexcellent super capacitive performance [36]. The groupcarbonized CoeMn-MOF-74 nanoparticles at 600e800 \\\\x03Cunder nitrogen flow for 2 h to obtained CoeCoOeMnO@Ccomposites. The schematic illustration for the preparationof CoeCoOeMnO@C composites is shown in Fig. 8. At apyrolysis temperature of 700 \\\\x03C, the porous electrode deliveredCo2+HydrothermalCrystalizationCoMn-MnanocrMn2+C8H6O6Fig. 8. Schematic illustration for the preparation of CoeCoOeMnO@C compoRef. [36], with permission from The Royal Society of Chemistry.a maximum specific capacitance of 800 F g\\\\x011 at a massnormalized current of 1 A g\\\\x011. The reasons for such good resultincluded the support gained from embedment of metal oxidesparticles in the carbon framework, more active contact areaprovided from the uniform dispersion of active metal oxidenanoparticles, and reduced transport length, faster electronstransport, and better chargeedischarge efficiency provided byporous carbon framework (from the original MOF framework).Given the effectiveness and relatively straightforward syn-thesis procedures, other researchers had also employed thismethod for the synthesis of various other metal/metal oxides.In 2017, Han's team fabricated NiO (from Ni-MOF) with areversible specific capacitance of 324 F g\\\\x011 at a massnormalized current of 1 A g\\\\x011 [37]. Comparatively, Wu andHsu embedded Ni nanoparticles in partially graphitic porouscarbon and found that it had much superior capacitivebehavior [38]. Optimization was carried out to increase elec-trical conductivity while maintaining large surface area. At theend, NiO nanoparticles synthesized from an optimum pyrol-ysis temperature of 800 \\\\x03C, was able to achieve a reversiblespecific capacitance of 886 F g\\\\x011 at a mass normalized currentof 1 A g\\\\x011. These well-dispersed NiO nanoparticles alsoprovided extra redox capacitance for Faradaic reactions.Meanwhile, Zeng and co-workers reported the room tem-perature synthesis of CeO2 from Ce-BTC (prepared fromcerium salts and benzene-1,3,5-tricarboxylic acid (BTC)) bytreatment of Ce-BTC in an alkaline solution ofKOH for 2 h [39].SEM images of both structures are shown in Fig. 9. After alkalitreatment, some of the original dumbbell-shaped from Ce-BTCremained, while some collapsed into half dumbbell-shapedDirectCarbonizationOF-74ystalsM/MO@Cnanocrystalssites and SEM image of CoeCoOeMnO@C composites. Reproduced from(a) (b)(c)2 μm 1μm1μm2 μm(d)Fig. 9. SEM images of (aeb) Ce-MOF and (ced) CeO2. Reproduced from Ref. [39], with permission from Elsevier.225M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245structures. Nevertheless, CeO2 had demonstrated a remarkablestability up to 10,000 cycles and a capacitive performance of235 F g\\\\x011. This specific capacitance of this CeO2 can also beincreased more than three-fold to 779 F g\\\\x011 with the addition ofK4Fe(CN)6 (redox-active agent) into the alkali KOH solution.Lastly, the electrochemical capacitive performance MOF-derived Cr2O3 e carbon composite was also studied byUllah's group [40]. The as-synthesized nanocompositesexhibited a specific capacitance of 291 F g\\\\x011 at a massnormalized current of 0.25A g\\\\x011 and a 95.5% long-term cyclingstability over 3000 cycles. The porosity and high surface area ofthe composites obtained using this approach had made thempromising electrode material for use as supercapacitors.Table 1Comparison of the performance of various nanostructures synthesized from MOFMaterials Doping Original MOF Specific sarea (m2Carbon Nitrogen ZIF-8 934Functionalized C Nitrogen ZIF-8 1140C@MWCNT e ZIF-8 569.3Co3O4 e ZIF-67 128Co3O4/NiCo2O4 Nickel ZIF-67 46CoS Sulfur ZIF-67 96.3Double-shelled CoS Hydroxide, Sulfur ZIF-67 110NiCo2S4 Sulfur, Nickel Co-MOF 72Co/MnO/CoO/C e CoMn-MOF-74 242.4NiO e Ni3(HCOO)6 34Ni@Graphitic C e Ni-BDC 140CeO2 e Ce-BTC 205Cr2O3 Carbon MIL-101 438In summary, despite their disadvantages such as lowerspecific capacitance and lower power density, MOF-derivedcarbon, metal oxides, and their composites have promisingapplications in supercapacitors. In general, ZIF-8 derivedcarbon has the largest surface area, whereas Co-MOF derivednanostructures exhibit the highest specific capacitance. Amore detailed comparison of the performance of variousnanostructures synthesized from MOF precursors as super-capacitors, which are discussed in this section, is summarizedin Table 1 and graphically presented in Fig. 10. The strategieswhich are crucial for improvement of supercapacitive perfor-mance could be summarized as: i) well dispersion of nano-particles, ii) presence of porous carbon framework (or hollowprecursors as supercapacitors.urfaceg\\\\x011)Massnormalizedcurrent (A g\\\\x011)Specificcapacitance(F g\\\\x011)CyclenumbersReference0.1 285.8 1000 [25]1 373 5000 [27]1.0 326.0 10,000 [26]12.5 1100 6000 [28]5 972 12,000 [29]10 ~1400 1000 [30]5 769.9 10,000 [31]10 765.6 10,000 [32]1 800 1000 [36]1 324 1000 [37]1 886 1000 [38]1 779 10,000 [39]0.25 291 3000 [40][28][29][30]1000120014001600[25][27][26][31] [32][36][37][38][39][40]02004006008000 2 4 6 8 10 12 14Specific capacitance (F g-1)Mass normalized current (A g-1)Fig. 10. Comparison of the performance of various nanostructures synthesized from MOF precursors as supercapacitors [denoted by their reference number].226 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245shells) which provide support to hold nanoparticles and largerchannels for easy transport of electrolyte ions, iii) addition ofnitrogen or other doping to improve electrical conductivity,and iv) presence of a conductive network which facilitate thetransport of electrons.2.2. Lithium ion batteriesLithium ion batteries (LIBs) are the most vital system forstate-of-the art rechargeable batteries. However, despite theirlong lifespan and ease of integration into portable devices,they are not able to meet the requirements for higher energyand power demand nowadays [41]. In the recent years, asidefrom graphite, the commercial negative electrode material,various nanostructured materials such as carbon nanotubes,graphene composites, transition metal oxides, etc. have beenextensively investigated for applications in LIBs. The quest forthese alternative materials is motivated by the limited capacityand energy density of graphite.Previous works have demonstrated that the incorporation ofnitrogen atoms into the carbon networks could facilitate theformation of stronger interaction between the nitrogen-dopedcarbon structure and the lithium ions [42,43]. In 2014, Zhengand co-workers synthesized NPC from the pyrolysis of ZIF-8under nitrogen flow [44]. The team manipulated the nitrogencontent to improve the electrochemical performance withoutsacrificing the structural stability of the NPC. At an optimizednitrogen content of 17.72 wt.% and carbonization temperatureof 800 \\\\x03C, the synthesized materials achieved a specific ca-pacity of 2132 mAh g\\\\x011 after 50 cycles at a mass normalizedcurrent of 100 mA g\\\\x011. This facile MOF-carbonization methodhad produced excellent capacity more superior than other NPCmaterials reported in Ref [45] and [43].Zn(NO3)2PVPDMF: ethanol = 5:3150 oC 6h OHOOOH++Fig. 11. Schematic illustration of the preparation process of the hollow porous ZnO/Besides carbon materials, MOFs have also been used asprecursors for the fabrication of metal oxides, transition metaloxides, metal/metal oxides, andmetal oxides/carbon compositesfor LIB negative electrode materials. Earlier in 2013, Yang'steam examined the synthesis of porous carbon-coated ZnOquantum dots from the controlled pyrolysis of Zn4O-MOF-5[46]. The as-synthesized material achieved a high reversiblecapacity of 1200 mAh g\\\\x011 at a mass normalized current of75 mA g\\\\x011. As such, this study revealed the importance ofoptimization of pyrolysis temperature, duration of heating, andthe heating rate, because these properties directly affected thecrystal size of the ZnOparticles, which in turn affected the rate oflithium ions diffusion and the overall electrochemical perfor-mance. ZnO quantum dots supported on graphene produced byMOF precursor had also shown better performance compared tothose synthesized using atomic layer deposition method, whichhad a recorded reversible capacity of 960 mAh g\\\\x011 at a massnormalized current of 100 mA g\\\\x011, reported in Ref. [47].Recently, Song and co-workers synthesized a hollow porousZnO distributed evenly on carbon nanocages at a moderatepyrolysis temperature of 500 \\\\x03C [48]. A schematic illustration oftheir preparation process is shown in Fig. 11. Given its pristinestructurewhich effectively accommodated the agglomeration ofZnO and volume expansion/shrinkage during discharge/chargeprocess, the prepared materials had demonstrated higher spe-cific capacity compared to cobalt-doped ZnO@C reported inRef. [49]. At a similarmass normalized current of 100mAg\\\\x011, ahollow ZnO@C had a reversible capacity of 750 mAh g\\\\x011 over100 cycles, whereas a cobalt-doped ZnO@C had a reversiblecapacity of 725 mAh g\\\\x011 over 50 cycles only.Furthermore, Gao and co-workers have attempted thesynthesis of ZnO@C using formic acid as an organic ligandinstead of the usual terephthalic acid (H2BDC) and tested its500 oC, 5 oC/minN2C nanocomposites. Reproduced from Ref. [48], with permission from Elsevier.227M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245performance as negative electrode materials for LIB [50].Results (Entry 5, Table 2) showed that it has better reversiblecapacity than those prepared from MOFs containing H2BDC.The authors attributed the excellent stability and lithium ionstorage properties to the carbon support and the appropriatecarbon content matrix from formic acid which suppressed theagglomeration and growth of ZnO nanoparticles duringcycling process. In 2016, Zou's group dispersed porous ZnOnanoparticles (derived from ZIF-8) on MWCNTs [51]. Whenusing methylimidazole as organic ligands and carbon nano-tubes as support, the resultant material exhibited a lowerreversible capacity of 419.8 mAh g\\\\x011 at a mass normalizedcurrent of 200 mA g\\\\x011, but it showed excellent rate capabilitywhen tested at various current rate between 100 and1000 mA g\\\\x011. Even when the mass normalized current wasincreased to 1000 mA g\\\\x011, the material was able to retain aspecific capacity of 326.8 mAh g\\\\x011.The aforementioned examples demonstrate that ZnO can besynthesized from various Zn-MOFs. By manipulating thechoice of organic ligands, pyrolysis conditions, and choice ofsupport, ZnO with different pore sizes and specific capacitycan be synthesized. Besides ZnO, ternary Zn-based oxideshave also generated much research because of their highspecific capacity, low cost, and environmental friendliness.Challenges remained as to improve their inherent poor elec-trical conductivity and large volume changes upon cycling. In2014, Zou's group fabricated a novel porous ternary Zn-basedoxides, ZnOeZnFe2O4@C hybrid structures which had one ofthe most outstanding electrochemical lithium storage perfor-mance reported so far [52]. At optimum conditions, a revers-ible capacity of 1060 mAh g\\\\x011 at 500 mA g\\\\x011 massnormalized current was achieved. This high rate performanceTable 2Comparison of the performance of various nanostructures synthesized from MOFMaterials Doping OriginalMOFPoresize (nm)Carbon Nitrogen ZIF-8 2ZnO(QD)@C e MOF-5 35Hollow ZnO@C e MOF-5 eCoeZnO@C Cobalt MOF-5 eZnO@C e Zn3(HCOO)6 30ZnO/MWCNTs ZIF-8 3.8e14.4ZnO/ZnFe2O4/C e MOF-5 7ZnO/ZnFe2O4 e ZnFe Prussianblue analogueeZnCo2O4 Zinc MOF-74 ~18Co3O4 nanorings e Co-NTCDACo-PTCDAeCo3O4/3D nickel foam e ZIF-67 eCoSe@carbon nanoboxes Selenium ZIF-67 2e20CoS2 hollow prisms Sulfur ZIF-67 2e10FeS2 @S-doped C Sulfur MIL-88-Fe 3e10NiP2@C Phosphorus Ni-MOF-74 4e30SnO2@C Tin (Sn) Cu-BTC 100CuO/3D graphene network e Cu-BTC 16CuO e Cu-BTC eAnatase TiO2 e MIL-125 ~1.25Anatase TiO2 e MIL-125 ~1.9was attributed to the near equal molar ratio of Zn to Fe, andthe incorporation of nanoparticles in carbon matrix with hol-low spaces which helped to relieve the volume effect andmaintained the electrical connectivity integrity. Recently, Cai'sgroup synthesized similar ZnO/ZnFe2O4 hybrid nanostructuresbut with ZnFe Prussian blue analogue and without a carbonsupport [53]. This unique structure had displayed a highreversible discharge capacity of 704 mAh g\\\\x011 after 200 cyclesat 200 mA g\\\\x011. The synergistic effect from the two well-distributed ZnO and ZnFe2O4 phases within the hybridnanostructures have contributed to this superior cycling sta-bility at high mass normalized current. Moreover, this studyhad proven the potential applications of various MOFs inpreparation of hybrid nanostructures with excellent electro-chemical performance as negative electrode materials.In addition to ZnO, Co3O4 is another metal oxides whichhas high theoretical capacities (890 mAh g\\\\x011) [54]. In 2017,Du and co-workers employed a controlled calcination of Zn-doped MOF-74 at 400 \\\\x03C to fabricate porous ZnCo2O4 [54].When compared with Co3O4, the obtained porous nano-structured ZnCo2O4 has demonstrated improved specific ca-pacity, cycling stability, and rate capability, owing to thestrong synergistic effect between Zn and Co. It was shown thatwith an additional doping step during MOF synthesis, thepyrolysis of MOF could yield metal/metal oxide compoundswith superior properties. Knowing that Co3O4 have limitedcycling stability and rate capability, Su's team has employed anew strategy to overcome these limitations. They synthesizedCo-MOFs with different organic linkers (1,4,5,8-naphthalenetetracarboxylic dianhydride (NTCDA) and per-ylene-3,4,9,10-tetracarboxylic dianhydride (PTCDA)) andthen treated them with organic amine solution beforeprecursors as negative materials for lithium ion batteries.Mass normalizedcurrent (mA g\\\\x011)Reversible capacity(mAh g\\\\x011)CyclenumbersReference100 2132 50 [44]75 1200 50 [46]100 750 100 [48]100 725 50 [49]100 955 50 [50]200 419.8 100 [51]500 1060 60 [52]200 704 200 [53]100 1243 80 [54]100 1370 30 [55]5000 976 500 [56]100 660 100 [57]1000 737 200 [58]100 1336.5 200 [59]50 656 50 [63]100 880 200 [64]100 409 50 [65]100 470 100 [66]10,000 127 1100 [67]1000 166 500 [68]228 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245calcination [55]. The enhanced lithium storage performance(1370 mAh g\\\\x011 reversible capacity at a mass normalizedcurrent of 100 mA g\\\\x011) of the obtained material has proventhat the synthesis method was feasible for the preparation andmodulation of metal oxides and their morphology. Besidesthat, the electrical conductivity of Co3O4 can also be improvedby coating the metal oxides on electro-conductive substrates.This has been demonstrated by Fang's group, who coatedporous Co3O4 on three-dimensional nickel foam to optimizeits structural and mechanical properties [56]. A schematicillustration of the synthesis process and SEM images of thematerial is shown in Fig. 12. The as-synthesized materialsustained a reversible capacity of 976 mAh g\\\\x011 at a high massnormalized current of 5000 mA g\\\\x011 for 500 cycles, puttingthis material as one of the negative electrode materials withsuperior cyclic stability reported in the literatures. Moreover,aside from coating porous Co3O4 on conductive substrates, theelectrical conductivity and structural stability can also beimproved by confining the particles within a hollow carbon-enriched outer shell. In 2016, Hu and co-workers pyrolyzedSe-doped ZIF-67 to obtain CoSe nanoparticles confined withinhollow carbon shell [57]. Benefiting from the structural pro-tection and pathways for charge transport provided by thehollow carbon matrix, the as-synthesized CoSe@C showedexcellent cyclability by retaining a 91.6% discharge capacity(a)(b)AdsorptionCobalt Ion2-methylimidazolePorous Co3O4/3DNF Co-MOCalcination10 μmFig. 12. (a) Schematic illustration of synthetic process of Co3O4/3D nickel foamRef. [56], with permission from Elsevier.(660 mAh g\\\\x011) after 100 cycles at a mass normalized currentof 100 mA g\\\\x011. On the other hand, Yu's group had employedan ion-exchange method to convert ZIF-67 MOF into CoS2hollow prisms [58]. They first developed a fast ion-exchangemethod to produce ZIF-67 hollow prisms, followed by a sul-fidation reaction to convert ZIF-67 into CoS4 bubble-likeparticles. A subsequent thermal treatment then producedcrystalline hierarchical CoS2 prisms with multilevel hollowinteriors. The unique structural features of hollow interior andultrathin shells had allowed the as-synthesized nanoparticles toperform remarkably as negative electrode material for LIBwith high rate capability and long-term cycling stability(reversible capacity of 737 mAh g\\\\x011 after 200 cycles at a massnormalized current of 1000 mA g\\\\x011).In 2016, Fe-based and Ni-based MOFs were used tofabricate derivatives of porous nanostructures for high per-formance LIBs. Pang and co-workers synthesized FeS2@S-doped C from MIL-88-Fe and sulfur powder [59]. The as-synthesized material had a rice skeleton which was mainlyinherited from the templating effect of the rice-like Fe-MOFprecursor. This rice-like morphology was also said to beroughly maintained at a MIL-88-Fe to S powder mass ratio of1: 2.5, in which over-sulfurization will destroy its morphology.Benefiting from the MOF derived heterogeneous porousstructure as well as the sulfur-doped carbon of this composite,(c)F/3DNFNucleationGrowthGrowthhybrid, (bec) SEM images of Co3O4/3DNF grown for 1 h. Reproduced from229M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245the FeS2@S-doped C had achieved a large reversible capacityof 1336.5 mAh g\\\\x011 after 200 cycles, which was more superiorthan other FeS2-based materials such as those reported in Ref[60e62]. In contrast, Li's group prepared nanostructuredNiP2@C via calcination of Ni-based MOF-74 with phosphorus[63]. The use of a MOF had retained the porous structures ofthe composite, whereas the introduction of porous carbonround the NiP2 nanoparticles had greatly enhanced the elec-tronic conductivity and stopped nanoparticles coalescencefrom Ostwald ripening. As a result, a reversible capacity of656 mAh g\\\\x011 after 50 cycles at the mass normalized current of50 mA g\\\\x011 was reported. Lastly, other examples of MOF-derived metal oxides which exhibited some electrochemicalproperties are SnO2@C (from Cu-BTC) [64], CuO octahedralon 3-D graphene (from Cu-BTC) [65], CuO octahedral (fromCu-BTC) [66], and porous anatase TiO2 (Ti-MIL-125) [67,68].All the examples discussed in this section and their electro-chemical measurements are tabulated in Table 2 and graphi-cally presented in Fig. 13.In summary, these MOF-derived nanostructures demon-strated high reversible capacities, rate capabilities, and cyclingstabilities more superior, or at least comparable with graphite(specific capacity ¼ 372 mAh g\\\\x011 [69]), the commercialnegative electrode material. The most widely studied nano-structures are MOF-derived zinc oxides. However, the highestreversible capacity was achieved by MOF-derived cobalt ox-ides. Overall, this two-step MOF-derived method is also costeffective, convenient, and can be extended to the fabrication ofother porous metal oxides with defined structures for otherapplications other than energy storage. Currently, improve-ments in terms of durability and energy density of LIBs arestill very much needed. These improvements are expected tobe done through more modifications of coating and doping.Lastly, while all the aforementioned examples in this sectionare centered on their performance as negative electrode ma-terials for LIBs, it is also worth mentioning that MOF-derivednanostructures could also be used as positive electrode mate-rials for lithium sulfur batteries [70,71], lithium-oxygen bat-teries [72,73], and lithium-selenium batteries [74].[44]150020002500[46][48][49][50][51[53[54][55][57][59][63][64][65] [66]0500100000101Reversible capacity (mAh g-1 )Mass normalizFig. 13. Comparison of the performance of various nanostructures synthesized frotheir reference number].3. Applications of MOF-derived nanostructures incatalysisIn order to meet stricter environmental standards, thechemical industry is always interested in the development ofchemical processes that produce the least amount of chemicalwaste. In principle, catalysts with good activity and selectivitywill help to increase the productivity and reduce the chemicalwaste produced. Therefore in the chemical industry, catalysisstudy is an important tool and will continue to be a key for thedevelopment of sustainable chemical processes [75]. Althoughhomogeneous catalysts have been successfully applied inmany industrial processes [76], however, they still suffer fromseveral intrinsic drawbacks of homogenous catalysts, such asdifficulty in recovery of the catalysts from the reaction mix-tures and poor thermal stability of the catalysts. In compari-son, heterogeneous catalysts offers the advantages of longercatalyst life, easier separation, and more efficient recycling,thus allowing energy conservations and an improved overallhandling and process control [77]. As a rapidly developingclass of porous materials, MOFs have continued to show theirpotential applications in the field of heterogeneous catalysis.The catalytic applications of the direct use of MOFs for avariety of organic reactions, photocatalysis, and tandem re-actions have been reviewed comprehensively in Ref [78e80].In addition to their applications in electrochemical energystorage (cf. Section 2.0), nanostructures prepared from calci-nation or pyrolysis of MOFs can also be used as chemicalcatalysts. Conventional synthesis methods of chemical het-erogeneous catalysts, such as solvothermal, hydrothermal, andwet-impregnation methods, are challenging in obtaining cat-alysts with uniform active sites and surface areas [81].Therefore, facile alternatives for new/improved catalyst com-pound with better long-term stability and better pore sizecontrol are of research interest. When compared to porousmaterials synthesized from conventional methods, MOF-derived nanostructures generally have improved catalytic ac-tivities and selectivities due to their larger surface areas (withwell-defined and interconnected pore system) and facilely][52]][56][58][67][68]000010001ed current (mA g-1)m MOF precursors as negative materials for lithium ion batteries [denoted by230 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245tailored functionality [12]. The following sections will presentthe recent developments of MOFs as precursors for the prep-aration of various functional materials, such as metal andmetal oxides, in heterogeneous catalysis applications,including electrocatalysis, photocatalysis, and for the pro-duction of fine chemicals.3.1. ElectrocatalystsElectrocatalysts shape the future of sustainable energytechnologies by enhancing the rates of electrochemical re-actions, such as oxygen reduction reaction (ORR), oxygenevolution reaction (OER), and hydrogen evolution reaction(HER) that take place on the surface of electrodes. Theirrespective half-equations are shown in Fig. 14. Due to the highcost of platinum metal, researchers are interested in devel-oping low cost non-precious metal (NPM) catalysts that couldmatch the performance of commercial platinum-based cata-lysts in terms of stability and activity. Some examples of NPMcatalysts that have been studied include metal oxides, carbonmaterials (graphene, nanotubes, N-doped carbon), carbidesand borides [82].Meanwhile, MOFs (as precursors) have been proved to besuitable in derivation of highly active NPM electrocatalysts[86e88]. Some the earlier works which motivated researchersto develop MOF-derived metal/metal oxide-carbon nano-hybrids for electrocatalysis was carried out by Chaikittisilpand his team in 2014 [89]. Their works included the synthesisof ZIF-9 derived Co3O4/C hybrid that has shown similar cat-alytic activity in ORR in comparison to the commercial Pt/Ccatalyst. Since then, there has been many works reported onCo-MOF derived electrocatalysts. Xia's group synthesizedZIF-67 of various crystalline sizes ranging from 300 nm to afew mm and studied the effect of crystalline sizes on thepreparation of Co@N-doped carbon polyhedrons and theirORR activities and stabilities [90]. After pyrolysis under argongas flow for 2 h (at a temperature of 750 \\\\x03C), the Co@N-doped carbon polyhedrons prepared from the smallest MOFcrystals exhibited the highest ORR activity. The ORR onsetand half-wave potential values obtained were 0.86 V and0.71 V vs. RHE, respectively. The protective carbon shellsFig. 14. Half-reactions for the water splitting reaction:around the cobalt nanoparticles has successfully increased thestability of the electrocatalyst, one that is even superior to thatof Pt/C in an acidic medium. The commercial Pt/C catalyst hassuffered from a drastic current loss of approximately 87% after15,000 s, whereas the as-synthesized Co@N-doped carboncatalyst exhibited only a lower attenuation of 58%. The teamhas then improved the preparation method to synthesize anovel CoeCo3O4@C mixed metal oxide nanoparticlesencapsulated in a highly ordered porous carbon matrix [91].When coated on glassy carbon (GC) electrode, these highlydispersed nanoparticles (diameter of 15e30 nm) recorded anincreased onset potential of 0.93 V and half-wave potential of0.81 V vs. RHE in ORR.Recently, bimetallic MOF structures were used to prepareORR electrocatalysts with a low catalyst loading and highcatalytic activity. You and co-workers reported the synthesis ofa CoeNeC polyhedron electrocatalyst by direct pyrolysis ofZneCo bimetallic MOF without any post treatment [92]. Anillustration of the carbonization of ZneCo bi-MOF is shown inFig. 15. At optimum conditions, an excellent onset potential of0.98 V (almost comparable to the 0.982 Vof Pt/C) and a half-wave potential of 0.871 V vs. RHE were obtained. Overall, thesynthesized structure had well dispersed Co nanoparticles(~9.5 nm) with excellent catalyst stability and comparableORR activity than commercial Pt/C tested in the same study.Shang and co-workers reported a synthesis method forCo,N-doped carbon which is slightly complicated but effectivein increasing the activity of the as-synthesized electrocatalysts[93]. They coated ZneCo bimetallic ZIF nanoparticles withmesoporous silica, followed by pyrolysis at a temperature of900 \\\\x03C, and removal of the silica shell by wet chemicaletching. The coating of mesoporous silica had offered pro-tection to the Co,N-doped carbon nanoframework by pre-venting their irreversible fusion and aggregation during hightemperature pyrolysis. The optimal as-synthesized materialshowed that it had a superior ORR catalytic activity perfor-mance (a onset potential value of 0.88 V vs. RHE) ascompared to commercial Pt/C of the same loading in alkalimedia reported in the same study.There are some reported application of metal phosphatesand mixed metal oxides as electrocatalysts for OER. A facileORR [83], OER [84], and HER [85] respectively.(a)(b)Zn2+ZnxCo1-x(MeIM)2bi-MOFZnxCo1-x(MeIM)2bi-MOFCo2+Co-N-C-xCarbonizationMeIMFig. 15. Illustration of carbonization of the Zn/Co bi-MOF to produce CoeNeC. Reproduced from Ref. [92], with permission from American Chemical Society.231M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245strategy was employed by He's team to synthesize NieCoeP/C nanoboxes by chemical treatment of ZIF-67 with Ni(NO3)2and annealed with NaH2PO2 under nitrogen gas flow [94].This phosphidation approach has yielded a NieCoeP/Ccatalyst with superior OER activity (with an onset potential of1.56 V vs. RHE). This is comparable to the performance ofother metal-phosphate-based catalysts such as MneCoeP(1.56 V vs. RHE) [95], CoeP/reduced graphene oxide (rGO)(1.57 V vs. RHE) [96], and Ni2P (1.52 V vs. RHE) [97]. Theresearchers have attributed the enhanced OER performance (ascompared to NieCoeP in the same study) to the existence ofcarbon which improved the charge transfer conductivity andthe nanoboxes' hollow structure which presented more elec-trolyte/electrode contact area for the electrochemical reaction.On the other hand, NieCo mixed oxides nanocages wereprepared by Han's group by thermally treated NieCo Prussianblue analogue nanocages in air [98]. The NieCo Prussian blueanalogue was first treated with ammonia at room temperatureto undergo structural evolution and transformation into a cage-like nanostructure. The complex 3D cage-like hollow andporous structure had provided a large electrolyte/electrodecontact area, to which an excellent OER performance wasachieved with an onset potential of 1.61 V vs. RHE.Furthermore, nanoparticles embedded in NPC has provedits ability to function as bi-catalysts for both ORR and OERreactions. In 2015, Li's team loaded ZIF-9 derived Co3O4nanoparticles onto N-doped graphitic carbon layer/MWCNThybrid [99]. By tuning the ratio of MWCNTs to ZIF-9, theelectrocatalyst has optimum ORR performance with an onsetpotential of 0.89 V and a half-wave potential of 0.81 V vs.RHE. Furthermore, this highly active catalyst also catalyzedOER with an onset potential of 1.50 V vs. RHE. These resultshave demonstrated the important role of MWCNTs inproviding active sites and influencing the electron-conductiveproperties of the hybrid materials. In 2016, Lu and co-workersfabricated metallic Co nanoparticles embedded in NPC layers(Co@NPC) through pyrolysis of Co-BTC crystals [100]. Thecarbon layered structures, with a relatively uniform distribu-tion of Co nanoparticles, exhibited a good ORR and OERbifunctional catalytic activity. For ORR, an onset potentialvalue at 0.88 V vs. RHE was obtained. This performance wascomparable to that of commercial Pt/C catalyst. As for OER,the potential value recorded was 1.61 V vs. RHE as comparedto the commercial RuO2 (1.50 V vs. RHE). Furthermore,Dong's team has adopted a similar MOF-derived approach toprepare Co-embedded NPC nanosheets [101]. They employedmontmorillonite (MMT) as a template to obtain an intercalatedlayered structure of ZIF-67 and MMT, and subsequentlyremoval of MMT by hydrofluoric acid etching to yield a Co-embedded porous carbon nanosheets. A schematic illustra-tion of this synthesis strategy is shown in Fig. 16. Thiseffective and novel strategy had produced MOF-derivednanostructures with comparable onset potential, half-wavepotential, density, and durability with respect to Pt/C elec-trodes tested in the same study.Similar to their applications in derivation of ORR and OERelectrocatalysts, ZIF-67 and Cu-BTC can also be used as pre-cursors to prepare promising catalysts for HER. The majorchallenges remain for HER electrocatalysts are low over-potential and poor stability. In 2015, Hou and co-workers py-rolyzed a mixture of graphene oxide and ZIF-67 to obtain alayered structured electrocatalyst of N-doped graphene/Co-doped C/N-doped graphene oxide [102]. The as-synthesizedstructures possessed excellent electrocatalytic activities forHER, OER, and ORR, with a low onset overpotential of 0.058 Vvs. RHE for HER, a small onset potential of 1.66 V vs. RHE forOER, as well as a potential of 0.97 V vs. RHE for ORR. Theenhanced performance of this electrocatalyst was attributed tothe larger surface area, doping of nitrogen, and the presence ofdual active sites provided by graphene oxide sheets.MMTMMTMMTSi O Al HIn MMT:Metal ionmodificationZIFintercalationHF etchingNanospace-confinedcarbonizationM/C@MMTZIF@MMTZIFMX+-MMTM-PCNFig. 16. Schematic illustration of the synthesis of the porous carbon nanosheets by ZIF intercalation, carbonization, followed by HF etching. Reproduced fromRef. [101], with permission from Elsevier.232 M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245The synergistic effects between various transition metalelements and the synergistic effects between these transitionmetals and heteroatom-doped carbon were studied by variousresearchers. In 2016, Chen's group developed NiCo bimetallicsulfide nanoparticles embedded in N,S-co-doped porous car-bon that showed electrocatalytic activity for HER with anonset overpotential of 98 mV vs. RHE, and a current density of10 mA cm\\\\x012 was achieved at an overpotential of 247 mV vs.RHE [103]. This HER performance is comparable to the mostactive NPM electrocatalyst reported to date. This compositecatalyst also exhibited excellent activities towards the OERwith an onset potential of 1.43 V vs. RHE. The researchersattributed the outstanding electrochemical performance to theuniformly distributed sulfides nanoparticles supported on highsurface area carbon, the protection of coreeshell carbon, andthe synergistic effect between metal sulfides and S,N co-dopedcarbon. Also in 2016, Ming and co-workers prepared a similarCoNi bimetallic electrocatalyst (CoeNieSe/C@Ni foam) butwith Se powder and Ni foam [104]. The team first coated ZIF-67 on Ni foam, then refluxed the sample with Ni(NO3)2$6H2Obefore heated under nitrogen atmosphere with Se powder.When tested on the electrocatalytic performance towardsHER, the as-synthesized sample needed an overpotential ofmerely 90 mV and 183 mV to achieve a current density of10 mA cm\\\\x012 and 100 mA cm\\\\x012 respectively (vs. RHE).Subsequent OER performance was evaluated and an onsetpotential of 1.51 V vs. RHE was achieved. The sample alsodemonstrated good stability by showing negligible decrease inpotentials after 24 h and 2000 CV cycles. The study haddemonstrated the effectiveness of the co-existing of NiSe2 andNi3Se4 in providing more active sites, the good conductivity ofthe NiSe2/Ni3Se4/C hydric nanomaterials, and the in situ for-mation of electrochemically active NiOOH phase at the sur-face which promoted the OER reaction. Besides that, anothertwo research works which evaluated both OER and HERbifunctional catalytic performance using MOF-derived elec-trocatalysts are reported in Ref [105,106]. Yu's group fabri-cated NieP porous nanoplates through low-temperaturephosphidation of NieNi Prussian blue analogue [105]. The as-synthesized material delivered a current density of10 mA cm\\\\x012 at a OER onset potential of 1.53 V vs. RHE,lower than that of Ni(OH)2 (1.59 V vs. RHE) and NiO (1.66 Vvs. RHE) obtained from NieNi Prussian blue analogue in thesame study. The electrocatalytic properties of NieP nano-plates was further investigated for HER activity in acidic so-lutions. An overpotential of only 110 mV vs. was required toachieve a current density of 10 mA cm\\\\x012. Benefiting from thisMOF-assisted synthesis method which enabled the incorpo-ration of carbon into nickel phosphides and the presence ofmetallic nickel phosphides with high electrical conductivity tofacilitate electron transfer. As a result, an electrocatalyst withexcellent electrocatalytic activity was synthesized. The phos-phidation strategy was also employed by Li and co-workers toobtain Co2P@N,P-doped C/CNT hybrid through pyrolysis ofCo3(PO4)2 MOF and further addition of CNTs by strongsonication treatment [106]. The resultant catalysts displayed ahigh onset potential of 1.51 V vs. RHE for OER and anoverpotential value of 0.154 V vs. RHE for the HER. ThisCo2P@N,P-doped C/CNT hybrid also maintained an excellentstability of up to 25 h when used as an electrocatalyst for bothanode and cathode in an overall water electrolyzer. Eventhough the stability was inferior to that of the well-establishedIrO2/Pt couple, the study had proved the effectiveness of Co2Pfor water splitting and the importance between the interactionof Co2P and carbon structure to promote its performance as abifunctional electrocatalyst.Moreover, the very first work on the growth of NiFebimetallic nanoparticles enveloped in N-doped graphenemicrotube for HER was demonstrated by Wu's team in the year2017 [107]. They proved that by controlling the pyrolysistemperature, the morphologies of N-doped graphene micro-tubes could be tuned into bamboo-like submicro-tubes andamorphous nanotubes. Although the HER performance ofthese bimetallic catalysts was low as compared to the com-mercial Pt/C electrode, however, they exhibited the highestHER activity at an overpotential of 0.071 mV vs. saturatedcalomel electrode (SCE) as compared to other NiFe nano-crystals [108] and single metal HER catalyst [109] reported todate. Motivated by the feasibility to control the thickness ofcarbon protection shell, Zhou and co-workers developed a 3-step strategy to synthesize CoSe2 bimetallic nanoparticlesembedded in defective carbon nanotubes [110]. They usedZIF-8 as precursor for the growth of defective carbon nano-tubes and ZIF-67 as the cobalt metal precursor. A schematic233M.H. Yap et al. / Green Energy & Environment 2 (2017) 218e245illustration for the synthesis is demonstrated in Fig. 17. TheHER activity testing revealed that the CoSe2@deffective car-bon nanotubes has an onset potential of \\\\x010.04 V vs. RHE.This same study has showed the importance of a core-shellstructure in enhancing the stability and current density of thecatalysts.Additionally, other MOF-derived metal-carbon compositewhich has demonstrated electrochemical properties for HERwas Cu/nanoporous carbon formed by the direct carbonizationof Cu-BTC. Raoof has prepared a Cu-BTC derived Cu/Ccatalyst which had an overpotential of approximately \\\\x010.8 Vvs. AgǀAgClǀKCl [111]. The same group also preparedbimetallic CuePt@nanoporous carbon with improved elec-trocatalytic activity for HER with an onset potential of\\\\x010.01 V vs. RHE [112]. Their works has also proven that thesynergistic effect of bimetallic nanoparticles was vital inenhancing the electrocatalytic activity of the nanomaterials.In summary, owing to their thermal and chemical stability,ZIF family, in particular ZIF-67 and ZIF-9, were widely usedas precursors for synthesis of electrocatalyst for ORR, OER,and HER. These ZIF-derived cobalt oxides exhibit superiorelectrocatalytic performance especially after metal doping andencapsulation by carbonaceous materials. A comparison of theperformance of various nanostructures synthesized from MOFprecursors as electrocatalyst for ORR, OER, and HER whichwere discussed in this Section 3.1 is tabulated in Table 3.Aside from the comparable onset potentials, long term sta-bility towards, ORR, OER, and HER (especially in acidicmedium) will be an important indicator in future research.Overall, the strategies which are crucial for improvementof electrocatalyst performance could be summarized as:Fig. 17. Schematic illustration of the synthesis procedure for CoSe2@defecti(i) prevention of aggregation and improvement of structuralstability by carbon or silica protection, (ii) presence ofcarbonaceous material to improve charge transfer efficiencyand conductivity, (iii) nitrogen or phosphorus doping to pro-vide additional active sites, (iv) int"}}},{"rowIdx":95,"cells":{"Researcher Name":{"kind":"string","value":"Jindong Chen"},"ORCID":{"kind":"string","value":"0000-0002-3940-9968"},"Researcher Query Keyword":{"kind":"string","value":"Noncentrosymmetric Pnictide NLO Materials"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Local Strain Engineering of Two-Dimensional Transition Metal Dichalcogenides Towards Quantum Emitters': 'Title: Local Strain Engineering of Two-Dimensional Transition Metal Dichalcogenides Towards Quantum Emitters\\\\nQuantum-dot-like states in molybdenum disulfide nanostructures due to the interplay\\\\nof local surface wrinkling, strain, and dielectric confinement\\\\nChristian Carmesin,1 Michael Lorke,1, 2 Matthias Florian,1 Daniel\\\\nErben,1 Alexander Schulz,1, 3 Tim O. Wehling,1, 2 and Frank Jahnke1\\\\n1Institute for Theoretical Physics, University of Bremen, P.O. Box 330440, 28334 Bremen, Germany∗\\\\n2BCCMS, University of Bremen, P.O. Box 330440, 28334 Bremen, Germany\\\\n3Present address: Leibniz Institute for Materials Engineering IWT, Badgasteiner Strae 3, 28359 Bremen, Germany\\\\nThe observation of quantum light emission from atomically thin transition metal dichalcogenides\\\\nhas opened a new field of applications for these material systems. The corresponding excited charge-\\\\ncarrier localization has been linked to defects and strain, while open questions remain regarding the\\\\nmicroscopic origin. We demonstrate that the bending rigidity of these materials leads to wrinkling\\\\nof the two-dimensional layer. The resulting strain field facilitates strong carrier localization due to\\\\nits pronounced influence on the band gap. Additionally, we consider charge carrier confinement due\\\\nto local changes of the dielectric environment and show that both effects contribute to modified elec-\\\\ntronic states and optical properties. The interplay of surface wrinkling, strain-induced confinement,\\\\nand local changes of the dielectric environment is demonstrated for the example of nanobubbles that\\\\nform when monolayers are deposited on substrates or other two-dimensional materials.\\\\nIn the past, single-photon emission has been realized\\\\nusing trapped-ion systems [1], NV centers in diamond [2],\\\\nand epitaxially grown quantum dots [3] with important\\\\napplications in quantum information. Recently, single-\\\\nphoton emission from spatially localized centers in tran-\\\\nsition metal dichalcogenides (TMDs) has been demon-\\\\nstrated [4–14]. For this purpose, TMD monolayers have\\\\nbeen placed over gold nanostructures [8, 9], substrates\\\\nwith etched holes [10], and arrays of dielectric micropil-\\\\nlars [11, 12]. In the latter case, up to 96% active single-\\\\nphoton emitters have been obtained. Quantum light\\\\nemission has been also achieved with nanobubbles, which\\\\nare formed in vertically stacked TMD structures [13, 14].\\\\nSingle-photon emission from atomically thin TMDs\\\\ncan originate either from structural defects resulting in\\\\nelectronic trap states, which are known to be a source\\\\nof photoluminescence centers [15, 16], or due to strain-\\\\ninduced potentials with three-dimensional carrier con-\\\\nfinements. In this paper, we show that the strain-induced\\\\noccurrence of localized electronic states is intrinsically\\\\nlinked to a non-continuous deformation of the TMD sur-\\\\nface. We find that bending of the atomically thin sheet\\\\nleads to formation of local surface wrinkles and bond de-\\\\nformations are identified as their microscopic origin. Due\\\\nto the strong dependence of the TMD band gap on local\\\\nstrain fields [17] this wrinkling induces a spatially local-\\\\nized carrier confinement. We show that this wrinkling\\\\ntakes place on length scales of 1 nm, providing efficient\\\\nthree-dimensional localized states and thereby enabling\\\\nsingle-photon emission. This supports recent experimen-\\\\ntal results [9], which first established the connection be-\\\\ntween surface wrinkling and localized emission, although\\\\nthe resolution of the used confocal microscopy is limited\\\\nto much larger µm length scales. The local wrinkling ef-\\\\nfect is different from surface corrugations on a 10−20 nm\\\\n∗ ccarmesin@itp.uni-bremen.de\\\\nscale, that is known for many 2D materials, as studied\\\\nfor graphene in Ref. [18].\\\\nTo investigate the interplay between TMD layer de-\\\\nformations and carrier confinement, we model a TMD\\\\nlayer with atomic resolution using a million-atom super-\\\\ncell. Based on valence force field simulations, new equi-\\\\nlibrium positions of the individual atoms in the bended\\\\nmaterial are determined. The information about the dis-\\\\nplaced atoms is used in a tight-binding electronic-state\\\\ncalculation for the supercell structure. For the example\\\\nof nanobubbles, we demonstrate that surface wrinkling\\\\nacts as the main cause for quantum confinement with\\\\nquantum-dot-like electronic states. The considered ge-\\\\nometry is depicted in Fig. 1(a). Nanobubbles are formed\\\\nwhen placing a monolayer of MoS2 on another MoS2\\\\nsheet used as a flat substrate, which is comparable to\\\\nrecent experimental investigations [13]. Starting from a\\\\nparaboloid profile of the nanobubble, the relaxation of\\\\natoms in the upper TMD layer is performed while keeping\\\\nthe atoms in the lower layer fixed. For this part, we uti-\\\\nlize a REAX potential [19] with the parametrization from\\\\nRef. [20] within a valence force field calculation, which\\\\nis capable of accurately describing bond deformations\\\\nunder strain as well as continuous bond formation and\\\\nbreaking dynamics. We find that both in-plane strain\\\\nand bending contribute to the shape of the nanobubble\\\\n(see Supporting Information for more details). Subse-\\\\nquent electronic-state calculations use this information\\\\nabout the positions of the individual atoms. For a su-\\\\npercell with an in-plane extension of 130 nm and up to\\\\n1.2×106 atoms, a 6-band tight-binding Hamiltonian is\\\\nsolved using the parametrization given in the Support-\\\\ning Information. Strain-induced local band gap changes\\\\narising from the displaced atomic positions are included\\\\nvia a generalized Harrison rule [21]. Additionally, when\\\\nlocally detaching the upper layer from the substrate un-\\\\nderneath and changing from a commensurate bilayer to\\\\nmonolayer-like structures across the nanobubble, a modi-\\\\nar\\\\nX\\\\niv\\\\n:1\\\\n90\\\\n2.\\\\n05\\\\n00\\\\n1v\\\\n1 \\\\n [c\\\\non\\\\nd-\\\\nma\\\\nt.m\\\\nes\\\\n-h\\\\nall\\\\n] \\\\n13\\\\n Fe\\\\nb 2\\\\n01\\\\n9\\\\n2h/r = 0.1 h/r = 0.15\\\\nh/r = 0.175\\\\n(b) Probability densities \\\\nelectrons\\\\nholes\\\\nE - EBL [meV] E - EBL [meV] E - EBL [meV]\\\\n(c) Spectra \\\\n0\\\\n10\\\\n20\\\\n30\\\\n0\\\\n10\\\\n20\\\\n30\\\\nnm\\\\nnm\\\\nEm\\\\nis\\\\nsi\\\\non\\\\n [a\\\\nrb\\\\n. u\\\\nni\\\\nts\\\\n]\\\\n0\\\\n2\\\\n4\\\\n6\\\\n8\\\\n0\\\\n2\\\\n4\\\\n6\\\\n8\\\\n× 105× 103\\\\n0\\\\n2\\\\n4\\\\n6\\\\n8\\\\n-500 -300 -500 -300 -500 -300\\\\nh/r = 0.1 h/r = 0.15 h/r = 0.175\\\\n0 1.5 x 10-3 0 2 x 10-3 0 5 x 10-3\\\\n0 2 x 10-3 0 1 x 10-2 0 5 x 10-2\\\\nh/r = 0.1 h/r = 0.15\\\\nh/r = 0.175\\\\n(a) MoS2 nanobubble\\\\nFIG. 1. (a) MoS2 air nanobubble with radius r and height h. (b) Top view of the probability densities for electrons and holes,\\\\nwhich are responsible for the strongest optical transition marked in (c). Results are shown for a bubble radius of r = 18 nm\\\\nand various aspect ratios h/r. (c) Emission spectra calculated from Fermi’s golden rule based on the tight-binding states for\\\\nthe aspect ratios in (b).\\\\nfied dielectric environment and electronic hybridization is\\\\nexpected [22–25]. We include both effects into our model\\\\nby changing individual tight-binding parameters based\\\\non GW calculations, as explained in the Supporting In-\\\\nformation. Based on the calculated tight-binding states,\\\\noptical emission spectra are obtained using Fermis golden\\\\nrule.\\\\nIn our analysis, we consider nanobubbles of different\\\\naspect ratios h/r and sizes. The degree of localization is\\\\ncontrolled by the aspect ratio, as shown in Fig. 1(b).\\\\nIn contrast, the nanobubble size does not change the\\\\nqualitative behavior, as presented in the appendix. For\\\\nh/r = 0.1 the single-particle states are delocalized along\\\\nthe edge of the bubble, whereas for an increasing aspect\\\\nratio of h/r = 0.15 distinct maxima are found. Due to\\\\nthe increased strain, these maxima carry the C3v sym-\\\\nmetry of the underlying crystal lattice. The trend of\\\\nstronger localization with increasing aspect ratio contin-\\\\nues for h/r = 0.175, which corresponds to the experimen-\\\\ntally reported value of Ref [13]. In this case, the proba-\\\\nbility densities are localized at three distinct positions at\\\\nthe edge of the nanobubble.\\\\nFigure 1(c) provides the emission spectra correspond-\\\\ning to the three different aspect ratios in Fig. 1(b), ob-\\\\ntained from all confined states of the TB model and\\\\nFermi’s golden rule. For h/r = 0.1, the oscillator strength\\\\nis very weak due to a small dipole interaction matrix el-\\\\nement, which is the result of a much larger radial broad-\\\\nening of the probability density for the electrons than\\\\nfor the holes. More optical transitions are present for\\\\nlarger aspect ratios h/r. In the case of h/r = 0.175\\\\nstrong emission from the electron and hole states, shown\\\\nin the right panel of Fig. 1(b), as well as a broad back-\\\\nground emission from other states is obtained. A sim-\\\\nilar behavior has been observed in recent experiments\\\\nwith WSe2 nanobubbles [14] and can be explained by\\\\nincreasing strain, which leads to a deeper confinement,\\\\nstronger localization of the single-particle states, as seen\\\\nin Fig. 1(b), and a redshift of the optical spectra.\\\\nTo analyze the mechanism behind the strong carrier\\\\nlocalization, we quantify the contributions to the con-\\\\nfinement potential due to strain as well as local changes\\\\nof the screening and electronic hybridization for the up-\\\\nper TMD layer. As a result of the nanobubble geome-\\\\ntry, shown in Fig. 2(c), the dielectric and electronic envi-\\\\nronment outside of the bubble corresponds to a bilayer,\\\\nwhile, in the center of the bubble, the large distance be-\\\\ntween top and bottom layer resembles the situation of a\\\\nmonolayer. This effect increases with increasing height\\\\nhmax of the bubble, as illustrated in Fig. 2(c). The re-\\\\nsulting changes of band gap are depicted in Fig. 2(d).\\\\nEffectively, it leads to a repulsive potential in the center\\\\nof the bubble. From our calculations we obtain band gap\\\\nchanges on the order of 100 meV, consistent with recent\\\\nmeasurements [24].\\\\nThe strain-induced changes of band gap for three dif-\\\\nferent aspect ratios are displayed in Fig. 2(a). By com-\\\\nparing these strain effects with the probability densities\\\\n3(a) Strain effect: Change of band gap\\\\n(b) Surface wrinkling: Strain pockets\\\\n(c) Nanobubble cross-section\\\\n(d) Dielectric effect: \\\\nChange of band gap\\\\nK\\\\nValence band \\\\ncharacter\\\\nPotential well\\\\n(nanostructure)\\\\n(1) no contribution of dxy \\\\n(2) contribution of dxy \\\\ndz2\\\\ndxy\\\\nK\\\\ndz2\\\\n(e) Orbital character of holes \\\\nwith strain and dielectric effect\\\\nε = ε\\\\nMoS2ε = ε\\\\nMoS2 h\\\\nmax ε = 1ε = 1\\\\nε = ε\\\\nMoS2 ε = εMoS2\\\\nα\\\\nh/r = 0.1 h/r = 0.15 h/r = 0.175\\\\nh/r = 0.1 h/r = 0.15 h/r = 0.175\\\\n0\\\\n10\\\\n20\\\\n30\\\\nnm\\\\n0\\\\n10\\\\n20\\\\n30\\\\nnm\\\\n0 0.04 0.08\\\\n-0.2 -0.1 0 -0.6 -0.3 0 -0.8 -0.4 0\\\\nFIG. 2. (a) Strain-induced band gap shifts for different aspect ratios h/r. (b) Average angle α (in radians) between normal\\\\nvectors of neighboring unit cells reveal wrinkling that corresponds to the spatially localized strain pockets. (c) Cross-section\\\\nthrough the nanobubble of height hmax. The geometric situation implies a change of the dielectric and electronic environment\\\\nfor the MoS2 layer that is forming the nanobubble. The arrows indicating an angle α between normal vectors of neighboring\\\\nunit cells on the discretized surface as used in (b). (d) Repulsive potential inside the nanobubble of height hmax = 34 A˚ and\\\\naspect ratio h/r = 0.175 as a result of the modified dielectric and electronic environment for the top layer. (e) Schematic\\\\nrepresentation of the orbital character of localized hole states. For shallow confinement, the states are composed of dz orbitals,\\\\nwhile for deeper confinement dxy orbitals also contribute.\\\\nin Fig. 1(b), we can conclude that the strain-induced\\\\nchanges of the band gap are responsible for the increasing\\\\nlocalization, as the dielectric effect is rotationally sym-\\\\nmetric (see Fig. 2(d)). For aspect ratios of h/r = 0.15\\\\nand h/r = 0.175, strain pockets are found. To illumi-\\\\nnate their origin, the average angle between normal vec-\\\\ntors at neighboring unit cells (schematically depicted in\\\\nFig. 2(c)) is displayed in Fig. 2(b), as described by the\\\\nnormal-normal correlation function (see Supporting In-\\\\nformation). These average angles reveal a local wrin-\\\\nkling of the surface for aspect ratios of h/r = 0.15 and\\\\nh/r = 0.175, with the angle of the normal vectors be-\\\\ntween neighboring unit cell varying up to 6◦. A compar-\\\\nison between Figs. 2(a) and (b) shows that the strain be-\\\\ncomes non-uniform and peaks in the wrinkles. The phys-\\\\nical origin of this behavior can be traced back to the high\\\\nYoung’s modulus of MoS2 (see e.g. Ref. [20, 26]). The\\\\nhigher the modulus, the better a material resists against\\\\ncompression/elongation. Especially for a high modulus,\\\\nthe material tends to minimize its total change of length.\\\\nIn particular, this is possible by the formation of wrin-\\\\nkles. The bond deformations responsible for the wrin-\\\\nkling are more prominent in TMDs than in graphene, as\\\\nthe former consist of a tri-atomic structure, in which the\\\\ntransition metal atoms are sandwiched between chalco-\\\\ngen atoms. Additionally, the spherical deformation of the\\\\nmaterial that is imposed by the nanobubble geometry fa-\\\\nvors wrinkle formation. We expect the wrinkles to be a\\\\ngeneral feature of spherical deformations, in contrast to\\\\ndepositing the material on a cylindrical surface or over a\\\\nridge. Mathematically, the difference between these situ-\\\\nations lies in the Gaussian curvature [27] of the geometry\\\\nthat is imposed upon the TMD monolayer. The strong\\\\nband gap variations, induced by strain and the dielec-\\\\ntric environment, also influence the oscillator strength of\\\\nthe optical transitions in Fig. 1(c). The main reason for\\\\nthis change is illustrated in Fig. 2(e). The localized elec-\\\\ntron states, originating from the conduction band, inherit\\\\nthe dz2 orbital character of the K-point. In the strained\\\\nbubble geometry, the Γ valley lies energetically higher\\\\nthan the K-valley for holes. Depending on the depth of\\\\nthe confinement potential, the orbital character of the\\\\nlocalized hole states is than either determined by states\\\\noriginating from the Γ-point (dz2), or from both the Γ-\\\\nand the K-point (dxy). In the former case, the emission\\\\nis weak, as the electron and hole wave functions carry\\\\nthe same orbital character (dz2). In the latter case, the\\\\nstrong dipole matrix element of the K-point transition\\\\ndominates, as known from the 2D monolayer [17]. These\\\\nlarge differences in the dipole matrix elements explain\\\\nwhy the measured recombination times of a few nanosec-\\\\nonds [4, 14] for localized states in TMD nanostructures\\\\nare much longer than the sub-picosecond range, which is\\\\nobserved for the 2D monolayer [28].\\\\nIn summary, our calculations provide insight into the\\\\nmechanism of quantum-dot-like carrier localization in\\\\n4TMD nanostructures for the example of MoS2 nanobub-\\\\nbles. We find a strong localization on the nanometer scale\\\\ninside strain pockets, which carry the C3v symmetry of\\\\nthe underlying crystal lattice. It is demonstrated that\\\\nthe formation of such strain pockets is caused by a non-\\\\nuniform strain, which peaks in surface wrinkles, an effect\\\\nthat is inaccessible to continuum theory. Our method is\\\\nbased on the requirements of covering strain as well as lo-\\\\ncal changes of the screening and electronic hybridization\\\\non the nanometer scale. We quantify the contributions\\\\nof these effects to the confinement potential in a com-\\\\nbined atomistic valence force-field and million-atom su-\\\\npercell tight-binding approach. Regarding the emission\\\\nproperties, it is shown that the orbital character of the\\\\nlocalized states changes compared to the 2D monolayer,\\\\nexplaining the measured recombination times of a few\\\\nnanoseconds [4, 14] in TMD nanostructures. We expect\\\\nthese properties and mechanisms to be present in other\\\\nTMD nanostructures, since the underlying physical prop-\\\\nerties are universal for TMD materials. Our results are\\\\nexpected to stimulate new and alternative realizations of\\\\nstrong carrier confinement in atomically thin TMD semi-\\\\nconductors.\\\\nACKNOWLEDGEMENTS\\\\nThe authors gratefully acknowledge computational re-\\\\nsources from HLRN Berlin and funding from the DFG\\\\n(”Deutsche Forschungsgemeinschaft”) via the graduate\\\\nschool ”Quantum-Mechanical Material Modeling”.\\\\nAppendix: Nanobubble geometry\\\\nStarting point of our relaxation procedure is a rectangular supercell of a MoS2 bilayer with periodic boundary con-\\\\nditions and an interlayer equilibrium distance of 3.0 A˚ in accordance with Ref. [20]. The top layer, which subsequently\\\\nforms the bubble, is shorter in both lateral directions so that its atoms are not influenced by the boundary conditions.\\\\nTo generate the bubble, the atoms of the top layer are displaced onto a paraboloid with a given height, corresponding\\\\nto the desired aspect ratio.\\\\nIn order to determine the relaxed atomic positions, we employ the LAMMPS code [29] using the REAX potential of\\\\nRef. [20]. To achieve a convergent geometry, we utilize a supercell of dimensions 130 nm× 130 nm× 15 nm, containing\\\\nabout 1.2 · 106 atoms. The relaxation is performed until a force tolerance of 10−3 eV/A˚ or a relative energy tolerance\\\\nof 10−16 is reached. In each iteration step, the charge equilibration energy is minimized until it reaches a value\\\\nbelow 10−3 eV. The obtained atomic positions provide access to height profiles of MoS2 bubbles for different aspect\\\\n(a) (b)\\\\nin-plane\\\\nbending\\\\nVFF\\\\n \\\\n \\\\nFIG. 3. Height profiles of MoS2 bubbles. (a) Curves for different aspect ratios. (b) Valence force field results for an aspect\\\\nratio of h/r = 0.2 and continuum elasticity results from E. Khestanova et al. [13] for dominantly in-plane strains (blue line)\\\\nand bending (red line).\\\\nratios h/r, which are depicted in Fig. 3(a). In Fig. 3(b), these values are compared to continuum elasticity theory\\\\n[13], where two cases, dominated by in-plane strain and bending, can be distinguished. Since our microscopically\\\\ncalculated height profile of the bubble is in between these limits, we conclude that both aspects contribute.\\\\n5ε0 -2.17187 eV\\\\nε2 -2.07972 eV\\\\nt0 -0.326583 eV\\\\nt0±2 0.561734 eV\\\\nt2 -0.411327 eV\\\\nt±2∓2 -0.355268 eV\\\\nt′0 -0.0537226 eV\\\\nt′0±2 -0.00459159 eV\\\\nt′2 0.052774 eV\\\\nt′±2∓2 -0.123627 eV\\\\nTABLE I. TB parameters for the MoS2 monolayer. Orbital energies and coupling matrix elements are denoted by ε and t,\\\\nrespectively. The primed terms describe next-nearest neighbor couplings.\\\\n1. Monolayer tight-binding model\\\\nTo describe the TMD monolayer, we employ an effective TB Hamiltonian with nearest and next-nearest neighbor\\\\ninteractions between Mo atoms and a {d0, d+2, d−2} basis set. The respective Hamilton matrix H in k-space is given\\\\nby\\\\nH =\\\\n\\\\uf8eb\\\\uf8ed Hd0d0 Hd0d2 Hd0d−2H∗d0d2 Hd2d2 Hd2d−2\\\\nH∗d0d−2 H\\\\n∗\\\\nd2d−2 Hd−2d−2\\\\n\\\\uf8f6\\\\uf8f8 , (A.1)\\\\nwith matrix elements\\\\nHαβ = 〈α, k|H|β, k〉 =\\\\n∑\\\\nR\\\\neik·R 〈α, 0|H|β,R〉 . (A.2)\\\\nFor the atomic positions R, we consider nearest neighbors R1, ..,R6 and next nearest neighbors R7, ..,R12, using the\\\\nin-plane coordinates\\\\nR1 = (1, 0), R2 = (1, 1), R3 = (0, 1),\\\\nR4 = (−1, 0),R5,= (−1,−1) R6 = (0,−1),\\\\nR7 = (2, 1), R8 = (1, 2), R9 = (−1, 1),\\\\nR10 = (−2,−1), R11 = (−1,−2) R12 = (1,−1)\\\\n(A.3)\\\\nin units of the the Bravais-lattice vectors\\\\na1 = a\\\\n(\\\\n1\\\\n0\\\\n)\\\\na2 = a\\\\n(−1/2√\\\\n3/2\\\\n)\\\\n(A.4)\\\\nwith the lattice constant a = 3.18 A˚. The matrix elements\\\\n〈α, 0|H|β,R〉 = fα,β(φ)tαβ(|R|) (A.5)\\\\ninclude the TB hopping parameter tαβ(R) towards nearest and next-nearest neighbors of the Mo-atoms, respectively.\\\\nWhile the threefold symmetry of the MoS2 lattice would be enhanced to a six-fold symmetry when considering the\\\\nMo sublattice only, we retain the original C3v symmetry by including different phase factors φ towards nearest and\\\\nnext-nearest neighbours\\\\nfα,β(φ) =\\\\n\\\\uf8f1\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f2\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f4\\\\uf8f3\\\\nnearest neighbors:\\\\nexp(−iα(φ− pi/6) + iβ(pi + φ+ pi/6)), if φ = 0, 2pi/3, 4pi/3\\\\nexp(−iα(φ+ pi/6) + iβ(pi + φ− pi/6)), if φ = pi/3, pi, 5pi/3\\\\nnext nearest neighbors:\\\\nexp(−iαφ+ iβ(pi + φ)), if φ = pi/6, 3pi/2, 5pi/6\\\\nexp(−iα(pi + φ) + iβφ), if φ = pi/2, 7pi/6, 11pi/6.\\\\n(A.6)\\\\n6Following Ref. [30], we include spin-orbit coupling according to\\\\nHSO =\\\\n[\\\\nH + λ2Lz 0\\\\n0 H − λ2Lz\\\\n]\\\\n(A.7)\\\\nwith\\\\nLz =\\\\n\\\\uf8ee\\\\uf8f0 0 0 00 0 2i\\\\n0 −2i 0\\\\n\\\\uf8f9\\\\uf8fb . (A.8)\\\\nThe Hamiltonian in Eq. (A.7) is block diagonal, which means that the z-component of the spin is not mixed by the\\\\nspin-orbit coupling and serves as a good quantum number due to the σh symmetry [30]. By using our TB approach, the\\\\nband structure of DFT calculations is well approximated, as shown in Fig. 4. The DFT calculations were performed\\\\nusing the VASP code [31, 32] together with the supplied projector augmented wave (PAW) potentials [33] and a 16 x\\\\n16 x 1 Γ-point sampling of the Brillouin zone. The used TB parameters for a MoS2 monolayer are given in Table. I.\\\\nFIG. 4. Tight-binding band structure. Left: Comparison of the three-band monolayer MoS2 TB band structure (red) with a\\\\nDFT calculation (black). Right: Character of the three bands dz2 (red), dx2 (green) and dxy (blue).\\\\nHamiltonian at points of high symmetry\\\\nBoth strain and changes of the dielectric environment influence the band gap of MoS2. To incorporate these effects\\\\ninto the TB model, we diagonalize the TB Hamiltonian analytically at points of high symmetry, which allows for an\\\\nidentification of the TB parameters to be adapted. For the Γ-point we find\\\\nH(Γ) =\\\\n\\\\uf8eb\\\\uf8edε0 + 6t0 + 6t′0 0 00 ε2 − 3t2 + 6t′2 0\\\\n0 0 ε2 − 3t2 + 6t′2\\\\n\\\\uf8f6\\\\uf8f8 (A.9)\\\\nwith the first diagonal element, denoting the eigenvalue of the valence band (VB) and the other two, describing the\\\\nconduction bands (CBs). At the K-point, we obtain\\\\nH(K) =\\\\n\\\\uf8eb\\\\uf8edε2 + 6t2 + 6t′2 0 00 ε0 − 3t0 + 6t′0 0\\\\n0 0 ε2 − 3t2 + 6t′2\\\\n\\\\uf8f6\\\\uf8f8 . (A.10)\\\\nIn the next section, we discuss how these results can be used to incorporate the effects of strain and of the dielectric\\\\nenvironment into the TB Hamiltonian.\\\\n2. Tight-binding model for nanobubbles\\\\nImplementation of strain\\\\nIn the case of a MoS2 monolayer, strain leads to an transition between a direct band gap at the K-point and an\\\\nindirect band gap between the Γ-point of the VB and the K-point of the CB, as demonstrated for instance in Ref. [34].\\\\n7From Eqs. (A.9) and (A.10), we obtain\\\\nECB(K)− EVB(Γ) = −9t0. (A.11)\\\\nHence, we can modify the nearest neighbor hopping parameter t0 to describe the direct to indirect transition in\\\\nstrained monolayers. We employ a generalized Harrison rule [21]\\\\nt0(r) = t0/ (1 + s)\\\\nη\\\\n(A.12)\\\\nwith s = (r − a)/a, where r is the inter-atomic distance and a the lattice constant. For an exponent of η = 11 the\\\\nexperimental band gaps in Ref. [35] are reproduced. The band structure for different strain situations within this\\\\napproach is depicted in Fig. 5.\\\\nFIG. 5. Band structure of a MoS2 monolayer for different strain situations s = (r − a)/a. The dashed lines indicate Fermi\\\\nlevels.\\\\nImplementation of the dielectric environment\\\\nIn addition to strain, the band gap can also be modified by changing the dielectric environment of the layer [23, 24].\\\\nWhile the system resembles a monolayer in the middle of the bubble, it corresponds to a bilayer at the edges. In this\\\\ntransition, not only the value of the band gap changes, but also its character. ARPES measurements [36] demonstrate\\\\nthat the MoS2 monolayer band gap is direct, whereas the bilayer band gap is indirect. Additionally, the value of the\\\\nband-gap at the K-point is reduced, due to more efficient dielectric screening. To determine this band gap EgK(z)\\\\nas a function of the interlayer distance z, we use data from DFT+GW calculations [17, 22]. Assuming a bowing-like\\\\nbehavior for the change of band gap with interlayer distance, a quadratic interpolation leads to\\\\nEgK(z) = 2.675 · 10−5z2 + 0.0021z + 2.3875 (A.13)\\\\nfor the direct band gap at the K-point.\\\\nTo introduce the influence of the air gap between the surface and bottom layer within the nanobubble into the\\\\nTB Hamiltonian, we follow the same route as in the previous section and determine the hopping parameter t0 as a\\\\nfunction of EgK(z). Thereby, the parameter t0 in our TB model depends on the interlayer distance according to.\\\\nt0(z) = −2\\\\n(\\\\nEgK(z)− (ε0 − ε2)\\\\n6\\\\n+ t2 + t\\\\n′\\\\n2 − t′0\\\\n)\\\\n. (A.14)\\\\n3. Normal-normal correlation\\\\nTo investigate the wrinkling of the material, we determine the normal-normal correlation function\\\\nNi =\\\\n1\\\\n3\\\\n∑\\\\nj,neighbors\\\\n~ni · ~nj (A.15)\\\\nthat measures the angle between the normal vectors of neighboring unit cells. Due to the symmetry of the system,\\\\neach cell i has three neighboring cells j, which gives rise to the normalization in equation (A.15). For better visibility\\\\nin Fig. 2b we depict the average angle between adjacent normal vectors αi = arccos(Ni).\\\\n84. Influence of nanobubble size\\\\nIn Fig. 6 we show the influence of both size and aspect ratio variations on the effective confinement potential for\\\\nthe nano-bubble. We find that the trend towards stronger localization is controlled by the aspect ratio, while the size\\\\nof the bubble controls the depth of the confinement potential and the distinctness of the observed localization.\\\\nAspect ratio\\\\nSi\\\\nze\\\\nh/r = 0.1 h/r = 0.15 h/r = 0.175 h/r = 0.2 Energy [eV]\\\\n−0.1 0 −0.2 0 −0.5 0 −0.8 −0.2\\\\n−0.15 0 −0.4 0 −0.75 0 −1 −0.5 0\\\\n−0.15 0 −0.4 0 −0.75 0 −1 0\\\\n−0.2 −0.05 −0.4 0 −1 −0.5 0 −1.5 −0.5\\\\nEnergy [eV]\\\\nEnergy [eV]\\\\nEnergy [eV]\\\\nFIG. 6. Effective confinement potential for aspect ratios of A=0.1,0.15,0.175, and 0.2 and nano-bubble radii of r=6, 12, 18,\\\\nand 24 nm.\\\\n9[1] D B Higginbottom, L Slodika, G Araneda, L Lachman, R Filip, M Hennrich, and R Blatt. Pure single photons from a\\\\ntrapped atom source. New Journal of Physics, 18(9):093038, 2016.\\\\n[2] A. Batalov, C. Zierl, T. Gaebel, P. Neumann, I.-Y. Chan, G. Balasubramanian, P. R. Hemmer, F. Jelezko, and J. Wrachtrup.\\\\nTemporal coherence of photons emitted by single nitrogen-vacancy defect centers in diamond using optical rabi-oscillations.\\\\nPhys. Rev. Lett., 100:077401, Feb 2008.\\\\n[3] Pascale Senellart, Glenn Solomon, and Andrew White. High-performance semiconductor quantum-dot single-photon\\\\nsources. Nature Nanotechnology, 12, Nov 2017.\\\\n[4] Ajit Srivastava, Meinrad Sidler, Adrien V. Allain, Dominik S. Lembke, Andras Kis, and A. Imamolu. Optically active\\\\nquantum dots in monolayer WSe2. Nat Nano, 10(6):491–496, 2015.\\\\n[5] M. Koperski, K. Nogajewski, A. Arora, V. Cherkez, P. Mallet, J.-Y. Veuillen, J. Marcus, P. Kossacki, and M. Potemski.\\\\nSingle photon emitters in exfoliated WSe2 structures. Nat Nano, 10(6):503–506, 2015.\\\\n[6] Chitraleema Chakraborty, Laura Kinnischtzke, Kenneth M. Goodfellow, Ryan Beams, and A. Nick Vamivakas. Voltage-\\\\ncontrolled quantum light from an atomically thin semiconductor. Nat Nano, 10(6):507–511.\\\\n[7] Yu-Ming He, Genevieve Clark, John R. Schaibley, Yu He, Ming-Cheng Chen, Yu-Jia Wei, Xing Ding, Qiang Zhang, Wang\\\\nYao, Xiaodong Xu, Chao-Yang Lu, and Jian-Wei Pan. Single quantum emitters in monolayer semiconductors. Nat Nano,\\\\n10(6):497–502.\\\\n[8] Johannes Kern, Iris Niehues, Philipp Tonndorf, Robert Schmidt, Daniel Wigger, Robert Schneider, Torsten Stiehm, Steffen\\\\nMichaelis de Vasconcellos, Doris E. Reiter, Tilmann Kuhn, and Rudolf Bratschitsch. Nanoscale positioning of single-photon\\\\nemitters in atomically thin WSe2. Adv. Mater., 28(33):7101–7105, 2016.\\\\n[9] Artur Branny, Gang Wang, Santosh Kumar, Cedric Robert, Benjamin Lassagne, Xavier Marie, Brian D. Gerardot, and\\\\nBernhard Urbaszek. Discrete quantum dot like emitters in monolayer MoSe2: Spatial mapping, magneto-optics, and charge\\\\ntuning. Appl. Phys. Lett., 108(14):142101, 2016.\\\\n[10] S. Kumar, A. Kaczmarczyk, and B. D. Gerardot. Strain-induced spatial and spectral isolation of quantum emitters in\\\\nmono- and bilayer WSe2. Nano Lett., 15(11):7567–7573, 2015.\\\\n[11] Artur Branny, Santosh Kumar, Raphal Proux, and Brian D. Gerardot. Deterministic strain-induced arrays of quantum\\\\nemitters in a two-dimensional semiconductor. Nature Communications, 8:ncomms15053, 2017.\\\\n[12] Carmen Palacios-Berraquero, Dhiren M. Kara, Alejandro R.-P. Montblanch, Matteo Barbone, Pawel Latawiec, Duhee\\\\nYoon, Anna K. Ott, Marko Loncar, Andrea C. Ferrari, and Mete Atatre. Large-scale quantum-emitter arrays in atomically\\\\nthin semiconductors. Nature Communications, 8:ncomms15093, 2017.\\\\n[13] E. Khestanova, F. Guinea, L. Fumagalli, A. K. Geim, and I. V. Grigorieva. Universal shape and pressure inside bubbles\\\\nappearing in van der waals heterostructures. Nature Communications, 7:ncomms12587, 2016.\\\\n[14] Gabriella D. Shepard, Obafunso A. Ajayi, Xiangzhi Li, X.-Y. Zhu, James Hone, and Stefan Strauf. Nanobubble induced\\\\nformation of quantum emitters in monolayer semiconductors. 2D Mater., 4(2):021019, 2017.\\\\n[15] Sefaattin Tongay, Joonki Suh, Can Ataca, Wen Fan, Alexander Luce, Jeong Seuk Kang, Jonathan Liu, Changhyun Ko,\\\\nRajamani Raghunathanan, Jian Zhou, Frank Ogletree, Jingbo Li, Jeffrey C. Grossman, and Junqiao Wu. Defects activated\\\\nphotoluminescence in two-dimensional semiconductors: interplay between bound, charged, and free excitons. Scientific\\\\nReports, 3:2657, 2013.\\\\n[16] Philippe K. Chow, Robin B. Jacobs-Gedrim, Jian Gao, Toh-Ming Lu, Bin Yu, Humberto Terrones, and Nikhil Koratkar.\\\\nDefect-induced photoluminescence in monolayer semiconducting transition metal dichalcogenides. ACS Nano, 9(2):1520–\\\\n1527, 2015. PMID: 25603228.\\\\n[17] A. Steinhoff, M. Ro¨sner, F. Jahnke, T. O. Wehling, and C. Gies. Influence of excited carriers on the optical and electronic\\\\nproperties of mos2. Nano Letters, 14(7):3743–3748, 2014. PMID: 24956358.\\\\n[18] Jannik C. Meyer, A. K. Geim, K. S. Novoselov, and T. J. Booth. The structure of suspended graphene sheets. Nature,\\\\n446:60, 2007.\\\\n[19] Adri C. T. van Duin, Siddharth Dasgupta, Francois Lorant, and William A. Goddard. Reaxff. a reactive force field for\\\\nhydrocarbons. The Journal of Physical Chemistry A, 105(41):9396–9409, 2001.\\\\n[20] Alireza Ostadhossein, Ali Rahnamoun, Yuanxi Wang, Peng Zhao, Sulin Zhang, Vincent H. Crespi, and Adri C. T. van\\\\nDuin. ReaxFF reactive force-field study of molybdenum disulfide (MoS2). J. Phys. Chem. Lett., 8(3):631–640, 2017.\\\\n[21] Sverre Froyen and Walter A. Harrison. Elementary prediction of linear combination of atomic orbitals matrix elements.\\\\nPhys. Rev. B, 20(6):2420–2422, 1979.\\\\n[22] Tawinan Cheiwchanchamnangij and Walter R. L. Lambrecht. Quasiparticle band structure calculation of monolayer,\\\\nbilayer, and bulk mos2. Phys. Rev. B, 85(20):205302, 2012.\\\\n[23] M. Ro¨sner, C. Steinke, M. Lorke, C. Gies, F. Jahnke, and T. O. Wehling. Two-dimensional heterojunctions from nonlocal\\\\nmanipulations of the interactions. Nano Lett., 16(4):2322–2327, 2016.\\\\n[24] Archana Raja, Andrey Chaves, Jaeeun Yu, Ghidewon Arefe, Heather M. Hill, Albert F. Rigosi, Timothy C. Berkelbach,\\\\nPhilipp Nagler, Christian Schller, Tobias Korn, Colin Nuckolls, James Hone, Louis E. Brus, Tony F. Heinz, David R.\\\\nReichman, and Alexey Chernikov. Coulomb engineering of the bandgap and excitons in two-dimensional materials. Nature\\\\nCommunications, 8:15251, 2017.\\\\n[25] Matthias Florian, Malte Hartmann, Alexander Steinhoff, Julian Klein, Alexander W. Holleitner, Jonathan J. Finley, Tim O.\\\\nWehling, Michael Kaniber, and Christopher Gies. The dielectric impact of layer distances on exciton and trion binding\\\\nenergies in van der waals heterostructures. Nano Letters, 18(4):2725–2732, 2018.\\\\n10\\\\n[26] Deji Akinwande, Christopher J. Brennan, J. Scott Bunch, Philip Egberts, Jonathan R. Felts, Huajian Gao, Rui Huang,\\\\nJoon-Seok Kim, Teng Li, Yao Li, Kenneth M. Liechti, Nanshu Lu, Harold S. Park, Evan J. Reed, Peng Wang, Boris I.\\\\nYakobson, Teng Zhang, Yong-Wei Zhang, Yao Zhou, and Yong Zhu. A review on mechanics and mechanical properties of\\\\n2d materialsgraphene and beyond. Extreme Mechanics Letters, 13:42 – 77, 2017.\\\\n[27] M. Nakahara. Geometry, Topology and Physics. CRC Press, 2017.\\\\n[28] C. Poellmann, P. Steinleitner, U. Leierseder, P. Nagler, G. Plechinger, M. Porer, R. Bratschitsch, C. Schu¨ller, T. Korn,\\\\nand R. Huber. Resonant internal quantum transitions and femtosecond radiative decay of excitons in monolayer wse2.\\\\nNature Materials, 14, 2015.\\\\n[29] Steve Plimpton. Fast parallel algorithms for short-range molecular dynamics. Journal of Computational Physics, 117(1):1–\\\\n19, 1995.\\\\n[30] Gui-Bin Liu, Wen-Yu Shan, Yugui Yao, Wang Yao, and Di Xiao. Three-band tight-binding model for monolayers of\\\\ngroup-vib transition metal dichalcogenides. Phys. Rev. B, 88:085433, Aug 2013.\\\\n[31] G. Kresse and J. Hafner. Ab initio. Phys. Rev. B, 47:558–561, Jan 1993.\\\\n[32] G. Kresse and J. Furthmller. Efficiency of ab-initio total energy calculations for metals and semiconductors using a\\\\nplane-wave basis set. Computational Materials Science, 6(1):15 – 50, 1996.\\\\n[33] G. Kresse and D. Joubert. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B,\\\\n59:1758–1775, Jan 1999.\\\\n[34] M. Ghorbani-Asl, S. Borini, A. Kuc, and T. Heine. Strain-dependent modulation of conductivity in single-layer transition-\\\\nmetal dichalcogenides. Phys. Rev. B, 87:235434, Jun 2013.\\\\n[35] Hiram J. Conley, Bin Wang, Jed I. Ziegler, Richard F. Haglund, Sokrates T. Pantelides, and Kirill I. Bolotin. Bandgap\\\\nengineering of strained monolayer and bilayer MoS2. Nano Lett., 13(8):3626–3630, 2013.\\\\n[36] Wencan Jin, Po-Chun Yeh, Nader Zaki, Datong Zhang, Jerzy T. Sadowski, Abdullah Al-Mahboob, Arend M. van der\\\\nZande, Daniel A. Chenet, Jerry I. Dadap, Irving P. Herman, Peter Sutter, James Hone, and Richard M. Osgood. Direct\\\\nmeasurement of the thickness-dependent electronic band structure of mos2 using angle-resolved photoemission spectroscopy.\\\\nPhys. Rev. Lett., 111:106801, Sep 2013.\\\\n', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAcc"}}},{"rowIdx":96,"cells":{"Researcher Name":{"kind":"string","value":"Huikang Jiang"},"ORCID":{"kind":"string","value":"-"},"Researcher Query Keyword":{"kind":"string","value":"Noncentrosymmetric Pnictide NLO Materials"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":97,"cells":{"Researcher Name":{"kind":"string","value":"Ning Ye"},"ORCID":{"kind":"string","value":"0000-0002-3679-4047"},"Researcher Query Keyword":{"kind":"string","value":"Noncentrosymmetric Pnictide NLO Materials"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"}}},{"rowIdx":98,"cells":{"Researcher Name":{"kind":"string","value":"Chao-Hui Zhang"},"ORCID":{"kind":"string","value":"0009-0009-5002-1890"},"Researcher Query Keyword":{"kind":"string","value":"Ultralight Integrated Anode for Lithium Metal Batteries"},"Research Full Paper":{"kind":"string","value":"{'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, as shown\\\\nin Fig. 13(b), which indicates that there is no nitrogen\\\\nin the ambient pressure recovered sample.\\\\nThe EDS spectra were measured one day after open-\\\\ning the cell but before the Raman spectrum in Fig. 9(e),\\\\ntherefore it should still be somewhat representative of\\\\nthe cubic sample. The WDS was measured on day 2, but\\\\nunfortunately, we do not know the state of the sample\\\\nbetween 1day and 5days after opening. Fig. 14 shows\\\\nthe Raman spectrum of the recovered sample used for\\\\nEDS measurements after 12days. We identify several16\\\\npeaks that do not originate from the carbon tape (the\\\\nsample could not be removed from the tape due to its\\\\nfragility). The background subtraction of this sample\\\\nis shown in Fig. 10(d), where we clearly see the reap-\\\\npearance of the peak at 240 cm−1which was not present\\\\nafter 5days; however, this could be due to the large in-\\\\ncrease in background between day 1 and day 5 making\\\\nthe 240 cm−1peak immeasurable. By day 12, the back-\\\\nground has decreased to a comparable value to day 1\\\\nand the 240 cm−1peak is observable again. The other\\\\npeaks after 5days could be present albeit less intense and\\\\nbroader so overall they are less distinct.\\\\nFIG. 14. Raman spectra of the recovered sample after mea-\\\\nsurements by EDS and WDS (12 days after opening). The\\\\nsample is on carbon tape whose Raman spectrum is shown\\\\nfor comparison. The stars are the modes identified as coming\\\\nfrom the sample. The subtracted spectrum is shown in Fig.\\\\n10(d).\\\\nS5: Raman tables of structures\\\\nHere we state the Raman and infrared excitations of\\\\nvarious space groups for LuH xstructures: cubic Fm3m,\\\\ntrigonal P3c1, hexagonal P63cm,P63, and P63/mmc,\\\\nand finally cubic Ia3for Lu 2O3andIa3-type. The ex-\\\\npected occupied Wyckoff positions for each space group\\\\nare written in table V which are then used to predict\\\\nthe number of excitations and their associated symme-\\\\ntries. The only clear expected selection rules concern\\\\ntheAmodes that are mainly Raman active in paral-\\\\nlel polarisations, except in hexagonal symmetries where\\\\nthey are Raman-active in both. For the Ia3-type phase\\\\nthat is proposed as the second phase at 1.9GPa, other\\\\nWyckoffpositionsshouldbeoccupiedbyhydrogenatoms.\\\\nUnfortunately, due to the low scattering cross-section of\\\\nhydrogen, we cannot determine which Wyckoff sites areoccupied. Calculations would be helpful to determine the\\\\nstoichiometry and the occupied sites.\\\\n[1] N. W. Ashcroft, Metallic hydrogen: A High-Temperature\\\\nS"},"Similar Full Paper":{"kind":"string","value":"{'Research Progress of Silicon/Carbon Anodes in Lithium-Ion Batteries': \"Title: Research Progress of Silicon/Carbon Anodes in Lithium-Ion Batteries\\\\nAbstract. People's need for energy is growing as science and technology advance, and finding effective ways to store and use energy has become crucial in today's world. Due to its effectiveness, environmental friendliness, and renewable nature, lithium-ion batteries (LIBs) are now a hot topic in study. However, traditional lithium-ion battery anode is usually graphite material, energy density limitations, its theoretical specific capacity is relatively low, difficult to meet the growing demand for high energy density. Moreover, graphite anode also has the disadvantages of limited multiplication performance, low first charge/discharge efficiency, and high charge/discharge platform voltage. Silicon-based materials have great potential for application in LIBs anode due to their high energy density, low de-embedded lithium potential, abundant resources, low cost, and good electrochemical properties. As a result, the materials based on silicon in LIBs are the main topic of this research. First, this paper, summarizes the advantages and challenges of the current silicon-based materials. Then, several forms of current silicon-based anode materials exist, including: silicon-carbon composites and alloying of silicon, are explored. Finally, improvement strategies for silicon-based materials are highlighted. 1 \\\\nIntroduction Nowadays, citizens can not live without lithium-ion batteries. Both the phones people use in the daily lives and the airplanes flying in the air are rely on the lithium-ion batteries to provide electricity. In recent years, electric cars are popular all around world, most of them are also use LIBs as their power sources. LIBs, as the most widely used energy storage tool today, are already changed peoples’ lives. One kind of secondary battery that may be charged and discharged frequently is the LIBs. The passage of lithium ions between the positive and negative electrodes is the primary factor governing its operation. Lithium ions are extracted from the positive electrode material during the charging process, transferred via the electrolyte solution to the negative electrode, and incorporated in the negative electrode material [1]. Concurrently, a current is created as electrons move across an external circuit from the positive electrode to the negative electrode. The battery is charged in this manner as the lithium in the positive electrode material progressively drops and the lithium in the negative electrode material gradually increases. Lithium ions separate from the negative electrode material during the discharge process, move through the electrolyte solution to the positive electrode once more, and embed into the positive electrode material [2]. Concurrently, electrons from the negative electrode go back to the positive electrode via an external circuit, creating a current that gives the device electrical energy. The battery discharges as a result of the progressive rise in lithium in the positive electrode material and the gradual reduction in lithium in the negative electrode material. Graphite is often used as the negative electrode material in lithium-ion batteries, whilst metal oxides containing lithium, such as lithium cobalt oxide and lithium manganese oxide, are used as the positive electrode material. Lithium ions are conducted between the positive and negative electrodes by the electrolyte solution [3]. Anode, as an important part of LIBs, deeply affects the specific energy of the battery. The cost of anode usually accounts for 5% to 15% of the total cost. Graphite, which is also known as carbon nanomaterials, is widely used as anode due to its superior performance in specific energy. An publication by Besenhard from 1975 described the first Li-ion intercalation based graphite electrode and demonstrated that graphite could intercalate several alkali-metal ions, including Li-ions [1]. Although the graphite anode has improved the safety of batteries, it also brings a serious limitation on battery’s energy density. And with the development of people’s life quality, the demand of the energy capacity is growing day by day. Scientists have to look for a new material of anode to replace the graphite and silicon based materials comes to their views [4]. The features, uses, and benefits of silicon-based materials utilized in LIBs are outlined in this article, which also offers recommendations for further lithium-ion battery research. EPEC 2024 , 01012 (2024)E3S Web of Conferences https://doi.org/10.1051/e3sconf/202455301012553 © The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0 (http ://creativecommons.org/licenses/by/4.0/). s 2 The advantages and challenges of silicon-based anodes Since one silicon atom can hold 3.75 lithium ions compared to one carbon atom in graphite, anodes made of silicon materials have a high theoretical specific capacity of 4200mAh/g, which can increase the battery's energy capacity by more than ten times when compared to anodes made of graphite. Furthermore, the silicon exhibits an operating voltage of 0.4 V vs Li/Li+, indicating its ability to inhibit lithium plating and offer increased safety. Without a doubt, silicon anode material is a great substitute for anode materials in the future [4]. Over 50 years have passed since silicon materials were first used in anodes. The usage of silicon as the anode material in lithium-ion batteries dates back to the early 1970s. Researchers led by Dey (1971) found that lithium could electrochemically alloy with Sn, Pb, Al, Au, Pt, Zn, Cd, Ag, and Mg at ambient temperature. Then, in 1976, two scientists named Sharma and Seefurth discovered that Li-Si alloys could function at a relatively high temperature of around 450°C. The Li22Si5 in particular has the largest theoretical capacity at 4200mAh/g. The first phase of research on silicon-based anodes took place between 1990 and 2000, during which time scientists mostly concentrated on the synthesis and utilization of silicon anodes. The electrochemical performance of the Si anode was thought to be improved by employing nano-sized Si composites with carbonaceous components. When Wilson, a scientist, combined nano-sized silicon with carbon matrix, the anode's capacity increased to 600mAh/g—nearly 50% more than that of commercial graphite [5]. Despite the silicon anode's many benefits, there are still several application-related challenges. The initial issue encountered during the silicon anode preparation process was the material's pulverization. Since a silicon atom can hold 3.75 lithium ions, the anode's volume will grow quickly as a result of the anode absorbing a lot of lithium. The battery's capacity will be lost and the silicon anode will be crushed as a result of the massive stress caused by the about 420% volume shift [6]. An further issue is that the significant volume shift will cause the solid-electrolyte interphase to become unstable. When the anode's potential falls below around 1 V against Li/Li+, a stable coating known as the SEI film forms on the silicon anode, preventing any more side chemical reactions. The silicon anode's electrochemical lithiation causes it to enlarge, which crushes the SEI coating. Due to an increase in ionic transport resistance, the exposed fresh electrode surface \\\\nresults in a thick SEI layer, which lowers electronic conductivity and coulombic efficiency. Furthermore, some active Li+ is consumed as a result of the SEI film forming when the SEI layer appears. This will lead to irreversible capacity and increased consumption of cathode electrodes [6]. 3 Research directions of several silicon anode materials After decades of hard work, scientists has achieved outstanding \\\\nresults on the application on silicon based anode and to some extent solved the problems mentioned above. They focus on improving the structure of silicon anodes and adding appropriate additives to enhance their performance. 3.1. Si/C composite materials Silicon-carbon (S/C) composites, as a new type of anode material in lithium-ion batteries, combine the advantages of both silicon and carbon, aiming at solving the problems existing in traditional anode materials, such as low specific capacity, large volume change, and poor cycling stability. The emergence of this composite material greatly improves the performance of lithium-ion batteries and is expected to promote their wider application. Silicon has poor electrical conductivity and a large volume effect during charging and discharging, which may lead to electrode polarization, material pulverization, SEI membrane reconfiguration, low Coulombic efficiency, and continuous capacity decay. Carbon material, on the other hand, has excellent electrical conductivity and mechanical properties, and composite with silicon can not only effectively alleviate the volume expansion, but also improve the electrode conductivity and obtain a stable SEI membrane. Silicon carbon composites are prepared by a variety of \\\\nmethods, such as sol-gel, mechanical ball milling, pyrolysis and high-energy ball milling. These \\\\nmethods enable silicon and carbon to be tightly combined to form anode materials with stable structure, good cyclability and high capacity. Currently, commercialized silicon-based anode materials are mainly composites of silicon oxide and nano-silicon with graphite, respectively, and the reversible capacity of silicon-carbon anode can be up to 450 mAh/g through the addition of silicon materials with a mass fraction of 5% to 10%.Although these materials can partially meet the application requirements in terms of Coulombic efficiency, cycling performance, and multiplicity performance, efforts are still being made to improve their energy density and cycling stability. The main advantage of silicon-carbon composites is their high theoretical gram capacity, which is about 4200 mAh/g or more, more than 10 times higher than that of traditional graphite-based anode materials. This means that if silicon-carbon composites can be successfully applied to lithium-ion batteries, the capacity and energy density of the batteries will be greatly improved. However, there are some challenges with silicon-carbon composites, such as volume expansion during charging and discharging and relatively short cycle life. In order to solve these problems, researchers are trying silicon powder nanosizing, silicon carbon coating, doping and other means, and have made some progress [7]. Figure 1 shows the typical S/C based anode for LIBs. EPEC 2024 , 01012 (2024)E3S Web of Conferences https://doi.org/10.1051/e3sconf/2024553010125532 Fig. 1. Typical S/C based anode for lithium-ion batteries [7]. For Si/C composite, there are some kind of choice of the carbon resources, such as polystyrene (PS), polyvinyl alcohol (PVA), polyvinyl chloride (PVC), etc. Catherine et al. used SiCl4 as a silicon source to undergo an oxidation-reduction reaction with KC8 to prepare a kind of Si/C composite based material. This material has been tested and proved that its initial discharge specific capacity is about 610mAh/g. After 10 cycles, the reversible specific capacity still have about 500mAh/g. And the graphite can not only composite with silicon, but also serve as a conductive skeleton or medium for other Si/C composite based materials [8]. Du et al. used triethanolamine to improve the composite material of graphite and nano silicon, which has a hollow structure and high reversible capacity [9]. 3.2. Silicon alloy Silicon alloy in lithium-ion batteries has many compelling advantages as a high-performance anode material. Silicon alloy not only inherits the high-capacity characteristics of silicon, but also improves some of the inherent defects of silicon through alloying, thus improving the overall performance of lithium-ion batteries. First, the high capacity of silicon alloy is one of its biggest highlights. Compared to traditional carbon anode materials, the specific capacity of silicon alloys can reach a much higher level, thus significantly increasing the energy density of the battery. This gives silicon alloy a significant advantage in application scenarios that require high energy density, such as electric vehicles and wearable devices. Second, silicon alloys effectively mitigate the volume expansion of silicon during charging and discharging through alloying. Silicon undergoes significant volume changes during charging and discharging, which can lead to the destruction of the electrode structure and rapid degradation of the battery capacity. The other metal elements in the silicon alloy can play a buffering role and reduce the volume change of silicon, thus improving the cycle stability and life of the battery. In addition, silicon alloys have good electrical conductivity. This helps to reduce the internal resistance of the battery and improve the charging and discharging rate and energy conversion efficiency of the battery. At the same time, the preparation process of silicon alloys is also being improved and optimized, making its production cost gradually reduced and more likely to achieve large-scale applications. However, the application of silicon alloys in lithium-ion batteries still faces some challenges. For example, the preparation process of silicon alloys is relatively complex, requiring precise control of the composition and ratio of the alloy. In addition, silicon alloys may still have certain capacity degradation problems during the cycling process, which requires further research and improvement. 4 Modification strategy 4.1. The application of binder Tding some binders is the solution to the silicon anode's expanding issue. The silicon anode can be fixed using self-healing binders. Traditionally, a polymer binder has been placed around the particles in the silicon anode in order to create connections between the active materials and current collector. The polymer binder and silicon particles fractured as a result of the silicon lithiating and increasing in volume during battery operation. Nonetheless, the self-healing binder is flexible and capable of spontaneously repairing the anode's mechanical devastation [6]. It has been discovered that carboxy group-containing polymers show promise as binders for silicon-based anodes. The polar hydrogen bonds hold the secret. Polar hydrogen bonds are formed between the binder and SiO2, and it is thought that these bonds have the ability to self-heal and regenerate when broken [10]. Fig. 2. (a) The lithiation/delithiation process of silicon particles in conventional adhesives and (b) SHPET adhesives [9]. As a binder for the silicon anode, Hao Chen's group has created a unique self-healing poly(ether-tioureas) (SHPET) polymer with a balanced stiffness and softness. The lithiation/delithiation process of silicon particles in SHPET adhesives and conventional adhesives is depicted in Figure 2. After extended cycles, SHPET is evaluated to ensure that it can restore the silicon anode that was harmed by significant volume fluctuations and to continue maintaining electrode integrity by constructing a stable electron transport channel and a low electrochemical resistance. After 250 cycles at 4200 mA/g, the Si/SHPET electrode has a high reversible EPEC 2024 , 01012 (2024)E3S Web of Conferences https://doi.org/10.1051/e3sconf/2024553010125533 capacity of 870mAh/g, indicating that it can sustain 85.6% capacity at such a high current density [11]. 4.2. Electrolyte additives As previously stated, an anode's stability in the SEI layer is crucial. Thick SEI layer development ultimately leads to low first CE and poor initial capacity. The electrolyte's constituents have a significant impact on the composition and shape of the SEI layer. Scientists have attempted to enhance the electrolyte's function in order to create a stable SEI layer. By mixing a number of chemicals into the electrolyte, such as vinylene carbonate (VC), tris(pentafluorophenyl)borane, succinic anhydride, and fluoroethylene carbonate (FEC), they attempt to combine the stable interfacial chemistry with integrated structural integrity [7]. By passivating the Si/electrolyte interphase, these additives can create a stable SEI. A less porous SEI layer covers the silicon anode in an EC/DEC electrolyte with 1.3M LiPF6, according to research by Nam-Soon Choi. In contrast, the SEI layer in an electrolyte without FEC tends to be porous and rough. The battery's performance also significantly improves; a half-cell with an electrolyte free of FEC has a progressive decline in coulombic efficiency after 40 cycles, but a half-cell with FEC-containing electrolyte maintains a coulombic efficiency of over 99% for up to 80 cycles [12]. 5 \\\\nConclusion In \\\\nconclusion, a great deal of study has been done by scientists to enhance battery performance, and the silicon-based anode was a popular subject. Even though silicon-based anodes have several benefits over conventional carbon-based anodes, there are still certain issues with their application. For example, the large volume change happened when the silicon is being lithiated, the resistance of electron transport increase due to the unstable SEI layers. Scientist solve these problem by mainly two kind of method, the first is improving the anode by making silicon based composite materials, the second is adding new kind of additives such as the electrolyte additives and the self-healing binders. composite materials made of silicon, such as Si/C composites and Si alloys. Because they have altered the structure of silicon-based anodes—for example, by producing a carbon shell to stop volume expansion during silicon lithiation—they are able to increase the performance of these anodes without rupturing the SEI layer or causing damage to the anode. Simultaneously, the anode's capacity may be increased by creating a composite material out of silicon and active materials to enhance electrode performance. Electrolyte additives can passivate the interface between the anode and electrolyte, making the SEI layer which formed on the silicon anode less porous and stable. The binders can repair the internal damage caused by the volume expansion of the silicon based anode. These two additives did not increase the performance of the electrode, but effectively solved two biggest problems of silicon-based anodes. Although these \\\\nmethods are highly effective, they can only be implemented in the laboratory and cannot be applied to large-scale production. The problem faced by future scientists is how to solve the problem of silicon-based anodes in a low-cost way, making it a safe and high-performance battery anode that surpasses traditional carbon based anodes, achieving large-scale commercialization, and replacing carbon based anodes. Although there will be many challenges in the future, from the content mentioned in this article, silicon-based anode materials are obviously a practical electrode material with huge potential and value of application. Future scientists may try to use silicon-based anode materials in complete battery design, achieving a shift from experimental exploration to basic applications.\", 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. "}}},{"rowIdx":99,"cells":{"Researcher Name":{"kind":"string","value":"Bo-Zheng Liu"},"ORCID":{"kind":"string","value":"0000-0002-4642-3688"},"Researcher Query Keyword":{"kind":"string","value":"Ultralight Integrated Anode for Lithium Metal Batteries"},"Research Full Paper":{"kind":"string","value":"{'Interfacial Characteristics of Na2FePO4F and Its Carbon Coated Material for Lithium/Sodium Hybrid Ion Battery': 'Title: Interfacial Characteristics of Na2FePO4F and Its Carbon Coated Material for Lithium/Sodium Hybrid Ion Battery\\\\nAbstract.................................................................................................................... XXI \\\\n1. \\\\nIntroduction.................................................................................................... 001 \\\\n 1.1 Overview of energy demand......................................................................... 003 \\\\n 1.2 Rechargeable batteries: a potential technology........................................... 005 \\\\n 1.2.1. Lead-acid batteries (Pb-acid)............................................................... 005 \\\\n 1.2.2. Nickel-cadmium batteries (Ni-Cd)....................................................... 007 \\\\n 1.2.3. Nickel-metal hybrid batteries (Ni-MH)................................................ 008 \\\\n 1.2.4. Redox flow batteries............................................................................ 009 \\\\n 1.2.5. Sodium sufur batteries (Na-S)............................................................. 010 \\\\n 1.2.6. Li-ion batteries (LIBs)........................................................................... 012 \\\\n 1.3 Na-ion batteries (NIBs): alternative technology for stationary \\\\napplications......................................................................................................... 015 \\\\n 1.3.1. Anode materials.................................................................................. 016 \\\\n 1.3.1.1. Carbons....................................................................................... 017 \\\\n 1.3.1.2. Titanium-based compounds....................................................... 017 \\\\n 1.3.2. Cathode materials............................................................................... 019 \\\\n 1.3.2.1. Layered oxides............................................................................ 020 \\\\n 1.3.2.2. Polyanionic compounds.............................................................. 022 \\\\n 1.3.3. Electrolyte............................................................................................ 024 \\\\n 1.4. Solid Electrolyte Interphase and Solid Permeable Interphase..................... 027 \\\\n 1.5. Aim of this doctoral thesis............................................................................ 031 \\\\n 1.6.', 'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pierre Toulemonde,1Gaston\\\\nGarbarino,2Patricia De Rango,1Sébastien Pairis,1Pierre Giroux,1and Marie-Aude Méasson1\\\\n1Institut Néel CNRS/UGA UPR2940, 25 Avenue des Martyrs, 38042 Grenoble, France\\\\n2European Synchrotron Radiation Facility, 71 Avenue des Martyrs, 38000 Grenoble, France\\\\n(Dated: Friday 1stDecember, 2023)\\\\nIn recent years, there has been a fervent search for room-temperature superconductivity within\\\\nthe binary hydrides. However, as the number of untested compounds dwindled, it became nat-\\\\nural to begin searching within the ternary hydrides. This led to the controversial discovery of\\\\nroom-temperaturesuperconductivityatonly1GPainnitrogen-dopedlutetiumhydride[Dasenbrock-\\\\nGammon et al., Nature615, 244 (2023)] and consequently provided much impetus for the synthe-\\\\nsis of nitrogen-based ternary hydrides. Here, we report the synthesis of stable trigonal LuH 3by\\\\nhydrogenating pure lutetium which was subsequently pressurised to ∼2GPa in a dilute-N 2/He-rich\\\\npressure medium. Raman spectroscopy and x-ray diffraction were used to characterise the structures\\\\nthroughout. After depressurising, energy-dispersive and wavelength-dispersive X-ray spectroscopies\\\\ncharacterised the final compound. Though our compound under pressure exhibits similar struc-\\\\ntural behaviour to the Dasenbrock-Gammon et al.sample, we do not observe any nitrogen within\\\\nthe structure of the recovered sample at ambient pressure. We observe two cubic structures under\\\\npressure that simultaneously explain the X-ray diffraction and Raman spectra observed: the first\\\\ncorresponds well to Fm3mLuH 2+x, whilst the latter is an Ia3-type structure.\\\\nINTRODUCTION\\\\nThe holy grail of room-temperature superconductiv-\\\\nity has been a long sought-after quest, ever since the\\\\ninitial predictions of superconductivity in metallic hy-\\\\ndrogen by Ashcroft in 1968 [1] and shortly after the\\\\npublication of BCS theory in 1957 [2, 3]. Though not\\\\npure hydrogen, many examples of high-temperature su-\\\\nperconductivity have been realised in recent years; these\\\\nhave reliably shattered high-critical-temperature (high-\\\\nTc) records with each new discovery. A notable example\\\\nwas SH 3with a Tcof 203K at 155GPa [4], as it provided\\\\ntangible promise for the field. Subsequent examples con-\\\\ntinued to push the threshold with the discovery of su-\\\\nperconductivity in YH 9and LaH 10at 243 and 260K re-\\\\nspectively both at approximately 200GPa [5–7]. Clearly\\\\nthesesuperconductingstatesrequireextremelyhighpres-\\\\nsures that few groups are able to reach, and this has been\\\\nthe primary technical challenge to overcome.\\\\nHence why the claim of room-temperature supercon-\\\\nductivity at 294K in nitrogen-doped (N-doped) lutetium\\\\nhydride at such a low pressure of 1GPa [8] has drawn\\\\nso much attention. Not only is it a new record Tcfor\\\\nsuperconductivity, but also it brings superconductivity\\\\nintothedomainofpracticablyachievableatnear-ambient\\\\nconditions. Furthermore, the samples are said to be\\\\nmetastable at ambient pressure which further adds to\\\\nthe wishful properties of such a material. In such a short\\\\nperiod of time, an impressive number of groups have al-\\\\nreadytriedtoreplicatetheresults, boththeoreticallyand\\\\nexperimentally [9–17], though a corroborative synthesis\\\\nremains elusive. Even Naturehas recently published an\\\\narticleentitled“Absenceofnear-ambientsuperconductiv-\\\\nity in LuH 2+xNy” by Ming et al.[14] in direct contentionwith the original Naturepublication [8], which goes to\\\\nshow how controversial this discovery has been.\\\\nN-doped lutetium hydride represents another step into\\\\nthe domain of ternary compounds following the exhaus-\\\\ntive hunt for binary hydride room-temperature supercon-\\\\nductors. This new domain is much larger and therefore\\\\nmore daunting to explore, so theoretical predictions are\\\\nvital to guide experimental works, and they have already\\\\nyielded several candidate compounds: Li 2MgH 16[18, 19],\\\\nYCaH 12[20, 21], ScYH 6[22]; and also the LaH 10-like\\\\nclathrate boronitrides La(BN) 5and Y(BN) 5[23]. Cal-\\\\nculations optimising superconductivity via doping have\\\\nalso shown that nitrogen from ammonia borane may af-\\\\nfect the superconducting properties of LaH 10[19, 24, 25].\\\\nExperimentally, the most notable confirmed example of\\\\na ternary hydride comes from Fm3m-(La,Y)H 10with a\\\\nsuperconducting temperature of 253K at 183GPa [26].\\\\nBeyond this, synthesising high-quality, high- Tcternary\\\\ncompounds under extreme pressures remains rare, thus\\\\nefforts that characterise this phase space in such extreme\\\\nenvironments are vital for the field.\\\\nIn order to synthesise N-doped lutetium hydride,\\\\nDasenbrock-Gammon et al.[8] and Cai et al.[27]\\\\nused pure lutetium with a H 2/N2gas mixture, whereas\\\\nother experimental papers started from pure lutetium\\\\nand NH 4Cl and CaH 2precursors [14, 15] which decom-\\\\npose to provide the required N 2and H 2. Here we choose\\\\nanotherprocess, byfirstsynthesisingpureLuH 3andthen\\\\nloading the diamond anvil cell (DAC) with a mixture\\\\nof dilute N 2and helium. We then methodically charac-\\\\nterise the obtained compound with Raman spectroscopy\\\\nand x-ray diffraction (XRD) at each step, and by x-ray\\\\nenergy-dispersive-spectroscopy (EDS) and wavelength-\\\\ndispersive-spectroscopy (WDS) at ambient pressure.arXiv:2304.04310v3 [cond-mat.supr-con] 30 Nov 20232\\\\nMETHODS\\\\nExperimental Methods\\\\nIn total we prepared three DACs with thin samples\\\\nof presynthesised LuH 3. Prior to synthesis, polished\\\\nlutetium metal was characterised by EDS and oxygen\\\\nand tantalum were observed in small quantities. The\\\\nLuH 3was then synthesised by hydrogen absorption us-\\\\ning the Sievert method by heating for 18hours at 200◦C\\\\nin 4MPa of H 2gas; further synthesis details are pro-\\\\nvided in the Supplementary Material (SM), Sec. S1\\\\n[28]. All samples came from this synthesis and were\\\\ndistributed amongst the three DACs. The first DAC\\\\n(DAC1) was loaded with a mixture of nitrogen and he-\\\\nlium, where we estimate that the quantity of N 2in the\\\\npressure chamber was 4nmol whilst the quantity of LuH 3\\\\nwas 11nmol. The other two DACs (DAC2 and DAC3)\\\\nwere loaded with nitrogen: DAC2 was loaded with a\\\\ngas loader, whereas DAC3 was cryogenically loaded with\\\\nliquid nitrogen. Amongst the DACs, only the sample\\\\nwithin DAC1 showed structural and chemical transfor-\\\\nmations under pressure which are discussed in the main\\\\ntext of this paper. The other DACs and further details\\\\nare discussed in the SM [28]. A ruby ball (for pressure\\\\nmeasurement) and a piece of silicon (for optimising the\\\\nRaman signal) were also placed inside the pressure cham-\\\\nber. DAC1 was sealed at 1.9GPa and characterised by\\\\nRaman spectroscopy and XRD. Though the sample was\\\\neventually heated to 65 °C at 1.9GPa, the main text only\\\\npresents data prior to heating, as heating had no effect\\\\non the structural properties.\\\\nThe XRD study was performed on the European\\\\nSynchrotron Radiation Facility (ESRF) ID15B beam-\\\\nline with λ=0.411Å at 300K. Polarised Raman scat-\\\\ntering was performed in quasi-backscattering geometry\\\\nat 300K with an incident laser line at 532nm from a\\\\nsolid-state laser. The scattered light was analysed by\\\\na single-grating spectrometer and a triple-grating sub-\\\\ntractive spectrometer; both were equipped with liquid-\\\\nnitrogen-cooled CCD detectors. We measured the Ra-\\\\nman signal of pure LuH 3just before loading in the DAC,\\\\nafter loading at 1.9GPa, before and after heating, and\\\\nfinally after returning to ambient pressure. After depres-\\\\nsurising, we analysed the composition of the sample with\\\\nEDS and WDS whilst primarily searching for nitrogen.\\\\nEXPERIMENTAL RESULTS\\\\nImaging of the sample\\\\nThecolourchangefromblueatambientpressuretored\\\\nat high pressure has been actively discussed in the litera-\\\\nture [8, 9, 15, 16]. Images of our sample in DAC1 before(300K, 1bar) and after (300K, 1.9GPa) loading are pre-\\\\nsented in Fig. 1. A white light was used to illuminate\\\\nthe sample in reflection and in transmission. Our LuH 3\\\\nsampleappearstranslucentwitharedcolourat1bar and\\\\nseems to become opaque at high pressure; however, this\\\\ncould be due to the majority of the sample rising up off\\\\nof the diamond during loading. After loading with the\\\\nmixture of He/N 2and pressurising to 1.9GPa, the sur-\\\\nface became reflective and blue. In Fig. 1c, we can also\\\\nsee a red region which remained flat against the diamond\\\\nwhich was also characterised and is discussed in Sec. S2\\\\nof the SM [28].\\\\nFIG. 1. White light images of the sample before [(a) and\\\\n(b)] and after [(c) and (d)] loading at 1.9GPa. Transmission\\\\nimages are shown in [(a) and (c)] and reflection images are\\\\nshown in [(b) and (d)].\\\\nX-ray diffraction\\\\nThe Rietveld fit of the XRD pattern measured on the\\\\ntrihydride in ambient conditions is shown in Fig. 2(a),\\\\nand we determine the structure to be trigonal P3c1with\\\\nlattice parameters of a=6.173(1)Å and c=6.424(1)Å.\\\\nThe lanthanide trihydrides tend to adopt either this\\\\ntrigonal structure or a hexagonal P63/mmcstructure\\\\n(the higher-symmetry parent group) [29]. Previously,\\\\nTkacz and Palasyuk [30] determined that LuH 3is hexag-\\\\nonal with a=3.57Å and c=6.41Å at ambient con-\\\\nditions. However, previous measurements had already\\\\nshown that the structure is trigonal with lattice parame-\\\\nters of a=6.16Å and c=6.44Å [31] which are similar to\\\\nour values. Furthermore, recent calculations by Dangić\\\\net al.predict that the trigonal structure should be more3\\\\nstablethanthehexagonalstructureinthispressurerange\\\\n[32]. Finally, the hexagonal structure would also be in-\\\\nconsistent with the Raman spectra we measured due to\\\\nhaving too few excitations, as shown in Table SIV of Sec.\\\\nS5 in the SM [28]. Overall we conclude that our start-\\\\ning LuH 3adopts a trigonal P3c1structure in ambient\\\\nconditions.\\\\nWith regard to impurities within our sample, from\\\\nthe Rietveld fit we determine that the sample is primar-\\\\nily LuH 3at 96.9(1) %, and the rest was identified to be\\\\nLu2O3. The Lu 2O3is likely to originate from deposits on\\\\nthe lutetium surface that were not removed by polishing\\\\nbefore hydrogenation. The space group of Lu 2O3isIa3\\\\nand the refined lattice parameter is 10.380(8) Å in agree-\\\\nment with the literature [33, 34]. We also show that the\\\\npercentageofLu 2O3staysconstantfor6monthswiththe\\\\nsample exposed directly to air (Sec. S2 of the SM [28]);\\\\nso the sample is stable with respect to oxidation within\\\\nthis time scale. The EDS measurements showed that a\\\\nsmall quantity of tantalum was present in the starting\\\\nlutetium; however, there are no signatures of tantalum\\\\nor tantalum hydride in the XRD spectra.\\\\nXRD patterns from the loaded sample at 1.9GPa are\\\\nshown in Fig. 2(b). They were measured in five differ-\\\\nent spots with sizes of 4 x 3 µm and separated by 20 µm\\\\nin a cross-shape. The results on the different spots are\\\\nremarkably similar and indicate that the sample is ho-\\\\nmogeneous in this region [see inset of Fig. 2(b)]. By\\\\ncomparing the XRD patterns, the transformation to a\\\\nnew phase is clear. In their paper, Dasenbrock-Gammon\\\\net al.determine the synthesised ambient pressure sample\\\\nto consist of two distinct Fm3mphases [8]: the major-\\\\nity LuH 3−δNϵ“A” phase (92.25 %of the sample) has a\\\\nlattice parameter of aA=5.0298(4) Å, whilst the lattice\\\\nparameter of the minority LuN 1−δHϵ“B” phase (7.29 %)\\\\nisaB=4.7529(9) Å [8]. Under pressure at 1.9GPa, we\\\\nobtainsimilarXRDpatternsthatcanbereasonablywell-\\\\ndescribed by two Fm3mphases. Our majority phase\\\\n(≈60%) has a lattice parameter of a1=4.990(6)Å, whilst\\\\nour minority phase ( ≈40%) has a lattice parameter of\\\\na2=5.145(2)Å. We note that our majority phase is the\\\\none with the smaller lattice parameter, but more discon-\\\\ncertingly we notice that the lattice parameters of both of\\\\nour phases are larger than those of Dasenbrock-Gammon\\\\net al.despite our compound being under pressure. A\\\\ntempting explanation might rely on the synthesis pro-\\\\ncess which starting from pure LuH 3would tend to pro-\\\\nduce compounds with higher hydrogen content that are\\\\ncloser to the trihydride with an expanded lattice.\\\\nInterestingly, after pressurisation there are some small\\\\nreflections that cannot be described by the refinement\\\\nusing two Fm3mphases. Moreover, there is a clear\\\\ninconsistency between the two Fm3mphases and the\\\\nRaman spectra, as shall be discussed in more detail later.\\\\nThis leads us to reconsider the structural composition,\\\\nand our analysis is in favour of one Fm3mstructure and\\\\nFIG. 2. Rietveld refinements of the patterns measured at the\\\\nESRF (beamline ID15B, λ=0.411Å) at 300K. (a) The trigo-\\\\nnal LuH 3sample at ambient pressure. (b) The high pressure\\\\ncompound at 1.9GPa and fitted with two Fm3mstructures,\\\\nstructures 1 and 2. Inset: patterns measured on five different\\\\nspots. (c) The high-pressure compound at 1.9GPa and fitted\\\\nwith one Fm3mstructure and one Ia3-type structure. Inset:\\\\nzoom of some of the weak reflections fitted by the Ia3-type\\\\nstructure (cf. arrows). Diff., difference between measured and\\\\ncalculated values.\\\\noneIa3structure.\\\\nIndeed, Fig. 2(c) shows that the small reflections can\\\\nbe better explained by refining the XRD data at 1.9GPa\\\\nwith one Fm3mstructure and one Ia3structure.\\\\nFrom this refinement, we obtained lattice parameters\\\\nof 4.99(3)Å and 10.329(3)Å for the Fm3mandIa3\\\\nstructures respectively. The lattice parameter of the\\\\nFm3mstructure remains the same within error as that\\\\nof the previous refinement using two Fm3mstructures.4\\\\nHere we exclude the presence of Fm3mLuH 3, since\\\\nthis phase was only observed previously above 12GPa\\\\n[30], far beyond our measured pressure range. However,\\\\nother Fm3mcompounds remain possible and shall be\\\\ndiscussed later.\\\\nRegarding the Ia3phase, we notice that it is similar\\\\nto the second Fm3mstructure but with an approximate\\\\ndoubling of the lattice parameter (2 a2, eight times the\\\\nvolume) and a slightly lower symmetry. Though the\\\\nIa3-type structure is similar to the Fm3mstructure,\\\\nthe lutetium atoms occupy different Wyckoff positions\\\\nwithin the lattice: namely the 8band24dsites. The\\\\n8bsite is highly symmetric, (1/4,1/4,1/4), whilst\\\\nthe24dsite is described by ( x,0,1/4) where xwas\\\\ndetermined to be approximately 0.975(8). This small\\\\ndifference from unity is indicative of a slight distortion\\\\nin the lutetium sublattice relative to the global cubic\\\\nsymmetry. The occupation of the 24dsite also has\\\\nramifications for the Raman activity as it provides eight\\\\nadditional phonons, whereas the 8bsite does not pro-\\\\nvide any. This shall be discussed further in later sections.\\\\nEventhoughthe Ia3phaseisreminiscentofLu 2O3, we\\\\nstate that it is not the same compound. Firstly, the lat-\\\\ntice parameter is smaller than the value of 10.357Å for\\\\nLu2O3at 1.9GPa, which was determined from the vol-\\\\nume dependence of Ref. [34]. Secondly, since the\\\\nIa3compound is recoverable (though metastable on the\\\\ntimescale of days as shown in Sec. S3 of the SM), we\\\\ndetermine that the ambient pressure lattice parameter is\\\\n10.41(1)Å (see Sec. S3 of the SM) which is larger than\\\\nthe ambient pressure value for Lu 2O3of 10.38Å [34]. To-\\\\ngether,theselatticeparametersatambientandhighpres-\\\\nsure indicate that the Ia3phase has a larger compress-\\\\nibility than Lu 2O3which further distinguishes them as\\\\nseparate compounds. Finally, the Raman spectrum, as\\\\nshown in the next section, does not contain the expected\\\\nmain Raman mode of Lu 2O3. Therefore, we conclude\\\\nthat the high-pressure sample of DAC1 does not contain\\\\ntwoFm3mphases, but in fact one Fm3mphase and\\\\noneIa3phase that we shall label as an Ia3-type phase\\\\nhenceforth.\\\\nRaman spectroscopy\\\\nWe first recall the nature of the Γ-point phonons ex-\\\\npected in the various space groups under consideration\\\\n(see Sec. S5 of the SM for more space groups [28]). From\\\\nthe literature on LuH 3(and YH 3), the crystal structure\\\\ncould correspond to Fm3morP3c1[29, 35, 36]. We ex-\\\\npect a total of 5A1g⊕12EgRaman active phonon modes\\\\ninthetrigonal P3c1phase,andasingleRaman-active T2g\\\\nmode in the Fm3mstructure, as stated in Table I. The\\\\nT2gmode is associated with the displacement of the hy-Space group Lu H1H2H3IR-active R-active\\\\nFm3m(LuH 3[12]) 4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2+x)4a8c4b- 2T1u 1T2g\\\\nFm3m(LuH 2[12]) 4a8c-- 1T1u 1T2g\\\\nP3c1(YH 3[36]) 6f2a4d12g6A2u+ 11Eu5A1g+ 12Eg\\\\nSpace group Lu1Lu2H1H2IR-active R-active\\\\nIa3(Ia3-type) 8b24d-- 7Tu 1Ag+2Eg\\\\n+5Tg\\\\nTABLE I. The total number of optical infrared-active (IR-\\\\nactive) and Raman-active (R-active) modes for the given\\\\nspace groups with the occupied Wyckoff positions stated for\\\\nvarious compounds.\\\\ndrogen atoms occupying the 8cWyckoff sites and is also\\\\nexpected to appear in Fm3mLuH 2andFm3mLuH 2+x.\\\\nHere we note that the Fm3mLuH 2and LuH 3are related\\\\nby the partial and continuous occupation of the octahe-\\\\ndral4bsites which results in the formation of LuH 2+x.\\\\nSpectroscopically and as shown in Table I, Fm3mLuH 3\\\\nand LuH 2+xbehave very similarly, whilst Fm3mLuH 2\\\\nlacks a T1umode since the 4bsite is completely unoccu-\\\\npied.\\\\nWide-range Raman spectra on the ambient pressure\\\\ntrigonal LuH 3and the high-pressure sample are shown\\\\nin Fig. 3(a). For the ambient pressure trigonal phase,\\\\nwe observe at least 12 features that are marked by\\\\nblack arrows. This is close to the 17 phonon modes\\\\nexpected for the trigonal P3c1structure and supports\\\\nour XRD analysis. Importantly, the number of modes\\\\nfar exceeds the four phonon modes predicted for the\\\\nalternative hexagonal P63/mmcstructure (see Sec.\\\\nS5 of the SM); so we can conclusively exclude it as a\\\\nviable structure. As we increase the pressure, we clearly\\\\nobserve the disappearance of all the phonons observed\\\\nassociated with the trigonal phase which is indicative of\\\\na structural transition. We also observe a large increase\\\\nin the background by a factor of ∼10, though we cannot\\\\nconclude whether it is intrinsic or due to the angle of the\\\\nsample as compared with the diamond. Most notably,\\\\nwe observe two peaks at high pressure that consistently\\\\nappear at approximately 1240 and 260 cm−1which were\\\\nnot present at ambient pressure.\\\\nAt energies below 260 cm−1we observe other features,\\\\nmost notably three weak excitations at 202, 164, and\\\\n128cm−1. As shown in Fig. 3(b), these are similar to\\\\nnot only those observed by Dasenbrock-Gammon et al.\\\\n[8] but also those osberved by Xing et al.[15], who as-\\\\ncribed them to vibrational modes of Fm3mcompounds.\\\\nHowever, the number of Raman modes is inconsistent\\\\nwith two Fm3mstructures, as we only expect one T2g\\\\nmode for each phase. Furthermore, we do not expect\\\\nthe lower-symmetry Wyckoff sites (e.g. 24e,32f, etc.)\\\\nto become occupied since hydrogen concentrations above\\\\nthee H atoms per Lu atom have not been observed at5\\\\nFIG. 3. (a) Raman spectra of trigonal LuH 3at ambient pres-\\\\nsure (blue) and a high-pressure sample at 1.9GPa (red). The\\\\ninset shows low-energy triple-stage data. (b) and (c) show\\\\nour data scaled on the Dasenbrock-Gammon et al.data at\\\\n∼2GPa [8]. We scale on the peak at 260 cm−1after a back-\\\\nground correction which aids the comparison. The scaling in\\\\n(b) is the same as in (c).\\\\nthese pressures. Herein lies the contradiction with these\\\\nprevious analyses: two Fm3mstructures cannot explain\\\\nthe number of phonon modes observed here and previ-\\\\nously [8, 15]. On the other hand, a distortion to a Ia3-\\\\ntype phase with lutetium atoms on the 24dWyckoff sites\\\\nprovides 1Ag⊕2Eg⊕5Tgphonon modes, and since the\\\\nlutetium atoms are heavy, these phonon modes would be\\\\nat low energy. Thus the Ia3-type phase could provide\\\\nthe required modes at low energy that were observed by\\\\nus and others [8, 15].\\\\nDISCUSSION\\\\nTo summarise the results, from the XRD we have ob-\\\\nserved a biphasic mixture of cubic Fm3mand cubic Ia3\\\\nby accounting for the numerous weak reflections. Theseweak reflections are not described by two Fm3mstruc-\\\\ntures. From the Raman spectroscopy, we observe one\\\\nstrong mode at 1240 cm−1and several weak modes at\\\\nandbelow260 cm−1. Thenumberofmodescannotbeex-\\\\nplained by two Fm3mstructures, whereas the Ia3struc-\\\\nture can in principle provide many modes at low energy.\\\\nAs clearly stated by Hilleke et al.[17], from the XRD re-\\\\nsults the identified sublattices of lutetium atoms (fcc for\\\\nanFm3mstructure and bcc for an Ia3structure) pro-\\\\nvides a constraint about which we should search but it\\\\ndoes not necessarily describe the entire structure. Now\\\\nwe shall discuss the possible origin of these structures,\\\\nand whether or not known compounds can explain the\\\\ndata.\\\\nFirstly, we shall address the contaminants which in-\\\\nclude Lu 2O3, pure tantalum, TaH 1−x, and the van der\\\\nWaals solid He(N 2)11[37]. This last compound forms be-\\\\nyondthepressurerangeofinterest(above9GPa)andthe\\\\nstoichiometry of the pressure medium is vastly different\\\\nfrom that of the compound, so we do not think that it is\\\\npresent. We have already shown that the Lu 2O3impuri-\\\\nties are minor in our XRD pattern at ambient pressure\\\\n(≈3%), so we do not expect a large effect from their pres-\\\\nence. Furthermore, wedonotseeanyRamansignatureof\\\\nthis phase. Indeed, the most intense Raman-active mode\\\\nof Lu 2O3is observed at 390 cm−1at ambient pressure\\\\n(shown in Sec. S3 of the SM [28]) and hardens slightly\\\\nup to 400 cm−1at 2GPa [34]. However, there is no in-\\\\ndication of this mode in any of the locations measured.\\\\nTherefore we eliminate Lu 2O3as being responsible for\\\\nthe XRD pattern and Raman-active modes, at either\\\\nambient or high pressure. Though the quantity is small\\\\n(≈1%), pure tantalum and TaH 1−xcould potentially be\\\\npresent. Pure tantalum forms an Im3mstructure [38],\\\\nwhereas TaH 1−xforms an I4m2structure [39]. Neither\\\\nstructure can explain the XRD reflections, and so we also\\\\neliminate pure tantalum and TaH 1−xfrom consideration.\\\\nOne should also consider intercalation effects from the\\\\npressure medium itself. Previous measurements have\\\\nshown that helium can occupy interstitial voids and\\\\nchange the structural properties of materials under pres-\\\\nsure [40–44]. This effect seems confined to network-\\\\nforming structures [40] or to materials possessing large\\\\nvoids such as single-wall carbon nanotubes [41, 42],\\\\nfullerenes [43], or clathrates [44]. However, neither trig-\\\\nonal, Fm3m, nor Ia3-type phases form these types of\\\\nstructures, and so we do not expect such helium interca-\\\\nlation; see Sec. S2 of the SM for further discussion. Nor\\\\nwouldweexpectanintercalationeffectfromN 2molecules\\\\ndue to their much larger size.\\\\nWewillnowcompareourXRDandRamanresultswith\\\\nthe known phases in the Lu-H-N landscape at room tem-\\\\nperature and ∼2GPa. These consist of pure N 2phases,\\\\nFm3mammonia (NH 3) [45, 46], fcc rock-salt LuN (RS-\\\\nLuN; NaCl-type B1,Fm3m), fcc zinc-blende LuN (ZB-\\\\nLuN;ZnS-type B3,F43m),hexagonalLuH δ(P63/mmc),6\\\\nand fcc LuH 2(CaF 2-type, Fm3m).\\\\nAt room temperature and 2GPa, pure N 2may form\\\\neither a fluid or a solid βphase. The β-phase crystallises\\\\nin aP63/mmcstructure [47, 48], and a single mode is\\\\nexpected at ∼2330cm−1, which we observe as a narrow\\\\npeakinthisrangeofenergy. N 2gashasnotonlyasimilar\\\\nvibron mode at high energy but also other peaks at low\\\\nenergy below 150 cm−1[49]. Some of the modes that we\\\\nmeasured might originate from N 2gas, but not the ones\\\\nat 195 and 166 cm−1or our dominant modes at 1240 or\\\\n260cm−1.\\\\nAmmonia could in principle form if hydrogen liberated\\\\nfrom the trigonal LuH 3lattice reacted with nitrogen in-\\\\nstead of being replaced by it. At 2GPa and ambient tem-\\\\nperature, ammonia is expected to form a Fm3mstruc-\\\\nture which should only possess one Raman-active mode\\\\n[46, 50]. Ammonia is unlikely to be detected by XRD\\\\ndue to the weak signal from the light atoms contrasted\\\\nagainst the large contribution from the massive lutetium\\\\natoms,thereforeitisunlikelythatanyoftherefinedcubic\\\\nphases could originate from it. Raman scattering under\\\\npressure shows that only modes at energies higher than\\\\n3100cm−1are observed in this phase [46]. So we exclude\\\\nammonia from being responsible for the Raman modes\\\\nwe measure at 1.9GPa.\\\\nThe primary potential nitride compound is Fm3mRS-\\\\nLuN which has a lattice parameter of a=4.7563(4)Å at\\\\nambient conditions [51]. Therefore this cannot explain\\\\neither of the two cubic phases observed by XRD, as the\\\\nlattice parameter will only continue to shrink under pres-\\\\nsure and it is already smaller than both of the lattice pa-\\\\nrameters measured. Furthermore, RS-LuN is in principle\\\\nRaman inactive since only the 4aand4bWyckoff sites\\\\nare occupied. Despite this, a strong excitation was ob-\\\\nserved previously at 582 cm−1and was ascribed to strong\\\\ndisorder [52]. Regardless, we do not observe this mode.\\\\nWe also note that the synthesis of RS-LuN is challenging\\\\nand previously required heating pure lutetium and nitro-\\\\ngen at 1600 °C [51]. Thus, since we have not laser-heated\\\\nour sample, we do not expect the formation of this com-\\\\npound. The EDS and WDS also support the idea that\\\\nRS-LuN did not form (see Sec. S4 of the SM) since this\\\\nwould result in a clear signature from nitrogen as this\\\\ncompound is stable at ambient pressure. On the other\\\\nhand, the F43mZB-LuN isomorph has only been pre-\\\\ndicted to form at pressures above 260GPa [53, 54]. Ex-\\\\nperimentally, the RS-LuN structure was shown to form\\\\npreferentially when synthesised at 30GPa and 2000K\\\\n[55]; that is to say, in far more extreme conditions than\\\\nwere attained here and in other papers, the ZB-LuN\\\\nstructure could not be formed, and so we do not con-\\\\nsider it viable from hereon.\\\\nSince we do not observe any signatures of trigonal\\\\nLuH 3and we do not expect cubic LuH 3at 2 GPa based\\\\non its predicted and observed stability [12, 17, 29, 30, 56],\\\\nit is likely that other lutetium hydrides have formed viathe decomposition of the trigonal LuH 3. Firstly, hexago-\\\\nnalP63/mmcLuH δcompounds(0 ≤δ≤0.2)formforlow\\\\nhydrogen concentrations [57–60]. At most, these hexag-\\\\nonal compounds could contribute four Raman-active\\\\nphonons which would help explain the low energy modes.\\\\nHowever, our attempts to reproduce the XRD patterns\\\\nwith any hexagonal structure at high pressure failed. We\\\\nnote that, in the recovered sample at ambient pressure,\\\\nwe were able to identify this phase (see Sec. S3 of the\\\\nSM).\\\\nThe other primary lutetium hydride is Fm3mLuH 2,\\\\nor the similar compound Fm3mLuH 2+xwith partially\\\\noccupied 4bsites. The lattice parameter of Fm3m\\\\nLuH 2is reported to be a=5.033 Å at ambient conditions\\\\n[14, 61, 62] which is also consistent with LuH 2+x. These\\\\nphases can therefore explain the XRD pattern of the re-\\\\nfined Fm3mphase. With regards to the Raman-activity,\\\\nwe expect one Raman-active T2gmode which was calcu-\\\\nlated to be between 960 and 1170 cm−1at ambient pres-\\\\nsure [32]. This would be consistent with the mode mea-\\\\nsured at 1240 cm−1at 1.9GPa. To explain our mode\\\\nmeasured at 260 cm−1, we note that an infrared-active\\\\nT1umode is predicted to appear at 250 cm−1inFm3m\\\\nLuH 3[12, 32]. Since Fm3mLuH 3and LuH 2+xare struc-\\\\nturally similar, one would expect that they share the pre-\\\\ndicted mode. LuH 2lacks this mode [32]. Thus, provided\\\\nthat the T1umode becomes Raman active, potentially by\\\\ndisorder, our excitations at 1240 and 260 cm−1could pro-\\\\nvide evidence for the presence of Fm3mLuH 2+x. Fur-\\\\nthermore, the blue colour observed in Fig. 1(d) would\\\\nalso be consistent with the formation of Fm3mLuH 2+x,\\\\nasitisalsopredictedtobeblue[63]. Insummary, Fm3m\\\\nLuH 2+xis consistent with both the Raman spectra and\\\\nXRDpatternswemeasured. However, itisclearthatthis\\\\nphase alone cannot explain the low-energy modes since\\\\nno other Raman-active modes exist, and the only other\\\\npredicted T1umode is at high-energy (above 1000 cm−1\\\\n[12, 32]).\\\\nThough we identify the Fm3mstructure as LuH 2+x,\\\\nwe still cannot explain the remaining Raman modes or\\\\ntheIa3phase identified by XRD results with known\\\\nphases. So, we shall discuss now the potential forma-\\\\ntion of the N-doped lutetium hydride compound. In Sec.\\\\nS3 of the SM [28], we show that once the pressure is\\\\nreleased, the sample is metastable but still contains the\\\\nFm3mandIa3phases. Most importantly, the recovered\\\\nsample does not contain nitrogen as shown by both the\\\\nEDS and WDS in Sec. S4 of the SM [28].\\\\nInfact, metalnitridesaregenerallychallengingtoform\\\\ndue to the significant activation barrier of the non-polar,\\\\ntriple-bondednitrogenatoms(bondenergy941kJmol−1)\\\\n[64]. However once synthesised, these nitrides tend to\\\\nhave refractory properties and are thermally and chem-\\\\nically stable [64]. Previously, Dierkes et al.synthesised\\\\nLuN by nitriding LuH 3[65], which is the closest analogy\\\\nto the desired reaction for this work. They note that ni-7\\\\ntridation does not start below 800 °C and even then the\\\\nuptake of nitrogen is slow until above 900 °C [65]; they\\\\nalso note that LuH 3begins to decompose by releasing hy-\\\\ndrogenabove300 °C.Perhaps, heatingwithinthiswindow\\\\nunder pressure would favour the formation of N-doped\\\\nlutetium hydride. Cai et al.performed a laser-heating\\\\nsynthesis at 1800 °C with pure lutetium and N 2/H2pres-\\\\nsure medium which formed a mixture of LuH 2and LuH 3\\\\nwith no observable nitride compounds [27]. Theoreti-\\\\ncally, it has been reliably noted that there are no ther-\\\\nmodynamically stable ternary Lu-H-N compounds: only\\\\nmetastable ones at best [10, 12, 17, 56, 66]. Furthermore,\\\\nwe prepared two pressure cells with pure nitrogen pres-\\\\nsure media and we observed no change in the trigonal\\\\nLuH 3structure upon heating to 65 °C at 2GPa followed\\\\nby pressurising to 12GPa. This indicates that nitrogen\\\\nhas a limited effect on the sample; further details are pro-\\\\nvided in Secs. S2 and S3 of the SM. So based on all of\\\\nthis, it would seem that the synthesis, as stated in the\\\\nNaturepaper [8], of heating the DAC for 24h at 65 °C\\\\nand 2GPa to form N-doped lutetium hydride would be\\\\nunlikely to occur.\\\\nFortunately, with the publication of Dias’ patent,we\\\\ncangaininsightintoanalternativesynthesismethod[67].\\\\nAccording to Fig. 1 of the patent, this patentable synthe-\\\\nsis involves heating lutetium metal in a reaction chamber\\\\nwith hydrogen and nitrogen gas at 4–10MPa and 200–\\\\n400°C for 12–24h before being pressurised to 3–20kbar\\\\nin a DAC [67]; this is rather different from the synthesis\\\\nstated in the Naturepaper [8]. Despite this, our syn-\\\\nthesis by pre-forming LuH 3at 200 °C with 4MPa of H 2\\\\nprior to loading is providentially similar, though we did\\\\nnot include nitrogen in this part of the synthesis. This\\\\npatentable synthesis is also very similar to the work of\\\\nDierkes et al.[65], though they did not heat with the two\\\\ngases together in the reaction chamber at the same time.\\\\nThiscombinedwithourworkstronglysuggeststhatheat-\\\\ning the pure lutetium metal in a hydrogen and nitrogen\\\\natmosphere at high temperatures (above 200 °C) is vital\\\\nfor the formation of the N-doped lutetium hydride.\\\\nOverall, these considerations for the nitridation of\\\\nlutetium hydride are also relevant for the partial or com-\\\\nplete nitridation of other rare-earth hydrides and for the\\\\nformation of other nitrogen compounds. Pragmatically,\\\\nthe successes of the rare-earth elements in producing\\\\nhigh-temperature superconductors and the prevalence of\\\\nammoniaboranesyntheseshavealreadyshiftedthedirec-\\\\ntion of research, as evidenced by the predictions of nitro-\\\\ngen doping of rare-earth compounds [19, 24, 25], or sim-\\\\nply rare-earth nitrogen compounds such as the clathrate\\\\nboronitrides La(BN) 5and Y(BN) 5[23]. As a result, the\\\\nincorporationofnitrogenintorare-earthhydridesisalog-\\\\nical route of inquiry for future experimental works where\\\\nthe challenges of nitrogen chemistry will have to be taken\\\\ninto account.\\\\nIn our case, we cannot conclusively say that we didor did not form N-doped LuH 3at 1.9GPa, as it could\\\\nhave decomposed and ejected the nitrogen prior to the\\\\nEDS and WDS measurements; however, it seems unlikely\\\\ngiven the arguments discussed. What is clear is that\\\\nat 1.9GPa, we formed a compound that is similar to\\\\nthat described by Dasenbrock-Gammon et al.[8], but\\\\nours was metastable and eventually decayed at ambient\\\\nconditions. What is also clear is that the contradictory\\\\nnature of observing many Raman-active phonons with\\\\ntwoFm3mlutetium lattices was an overlooked problem.\\\\nOverall, the question then becomes, what is the origin of\\\\ntheIa3-type phase?\\\\nTo explain the origin of the Ia3-type phase, we specu-\\\\nlate that this structure arises from a charge-density-wave\\\\n(CDW) distortion of a pure lutetium hydride compound.\\\\nPrevious work on the chemically similar ScH 3and YH 3\\\\nshows that there is an intermediate region between the\\\\nambient pressure trigonal or hexagonal structure and the\\\\nhigh-pressure cubic phase [68–70]. Theoretical work on\\\\nYH 3predicts that a Peierls distorted C2/mstructure\\\\nforms within this intermediate phase that continues to\\\\npossess a close approximation of a cubic sub-lattice [69].\\\\nUnfortunately, we tried an XRD refinement of the pro-\\\\nposed C2/mstructure without success, but this does not\\\\neliminate the possibility that this mechanism gives rise to\\\\nother distorted structures. A similar intermediate phase\\\\nwas also observed in ScH 3between 25 and 46GPa [70]\\\\nwhereas this phase was observed in YH 3between 9 and\\\\n24GPa [68]. Since lutetium is chemically similar to scan-\\\\ndium and yttrium, one could hypothesise that a similar\\\\nintermediate Peierls-distorted/CDW phase could arise in\\\\nour lutetium hydride compound. The CDW then pro-\\\\nvides a mechanism to form our Ia3-type phase which is\\\\nthen a distortion of a higher-symmetry phase; perhaps\\\\nFm3mdue to the already existing similarities. Further-\\\\nmore, the pressure range of the intermediate phase seems\\\\nto decrease with increasing atom size; that is to say, this\\\\nintermediate phase could then coincide with our mea-\\\\nsuredpressurerange. Itisalsoworthnotingthatastrong\\\\nchange in the optical gap has been observed within the\\\\nCDW phase in both YH 3and ScH 3[68, 70]. As such,\\\\nthe observation of poor-metal behaviour and upturns in\\\\nthe resistivity in previous measurements on lutetium hy-\\\\ndrides [14, 16, 71, 72] could then be evidence of a CDW\\\\nphase as the gap opens. Overall, a CDW phase driving\\\\nthe formation of the Ia3-type phase could then simul-\\\\ntaneously explain some of the electrical properties ob-\\\\nserved, the cubic lattice of lutetium atoms, and the forest\\\\nof Raman-active modes observed at low-energy without\\\\ninvoking the synthesis of a ternary compound.\\\\nCONCLUSION\\\\nWe obtain a biphasic sample which presents structural\\\\nsimilarities to the sample of Dasenbrock-Gammon et al.8\\\\n[8] by starting from pure trigonal LuH 3loaded in a DAC\\\\nat 1.9 GPa with a mixture of N 2/He. From x-ray diffrac-\\\\ntion, we clearly see a structural transformation from the\\\\ninitial trigonal phase to a mixture of cubic phases un-\\\\nder pressure. Similarly, with Raman spectroscopy we\\\\nobserve the loss of the modes associated with the trig-\\\\nonal structure and see the appearance of a strong mode\\\\nat 1240 cm−1that we associate with the T2gRaman-\\\\nactive mode of a cubic Fm3mstructure. However, we\\\\n(and others) observe more excitations than are possible\\\\nfor two Fm3mcubic structures. Overall we believe that\\\\nit is unlikely that these excitations come from impurity\\\\nphases since either they are not visible in XRD, they are\\\\nchemicallyunlikelytoform, orsimplytheirexcitationsdo\\\\nnotoccurintheenergyrange. Thusweconcludethatour\\\\nsample is a biphasic mixture of Fm3mLuH 2+xand an\\\\nIa3-typestructure, alsocomposedoflutetiumandhydro-\\\\ngen, which together may describe the XRD patterns and\\\\nRaman spectra. We postulate that the Ia3-type struc-\\\\nture is a distortion of a higher symmetry structure and\\\\ncould originate from a CDW phase. However, further\\\\ntheoretical work will be needed to support the origin and\\\\nstabilityofthisphase. Morebroadly, ourdiscussionofni-\\\\ntrogen chemistry will aid future works in experimentally\\\\nfinding ternary compound superconductors.\\\\nACKNOWLEDGMENTS\\\\nThis work is supported by the European Research\\\\nCouncil (ERC) under the European Union’s Horizon\\\\n2020 research and innovation program (Grant Agree-\\\\nment No 865826). This work has received funding from\\\\nthe Agence Nationale de la Recherche under the project\\\\nSADAHPT. We thank Abdellali Hadj-Azzem and Elise\\\\nPachoud for lutetium preparation, and Céline Goujon for\\\\nhelp in the preparation of the laboratory high-pressure\\\\nXRD setup. We thank Laetitia Laversenne for fruitful\\\\ndiscussions and Eva Zurek for stimulating exchanges of\\\\ninformation.\\\\nCOMPETING INTERESTS\\\\nThe authors declare no competing interests.\\\\nSUPPLEMENTARY MATERIAL\\\\nS1: Synthesis and techniques\\\\nLutetium (Alfa 3N) was characterised by EDS before\\\\npolishing it, whereupon oxygen was clearly identified in\\\\nLu2O3deposits with atomic concentrations between 20-\\\\n50%. A small amount of tantalum was also identified\\\\nas shown in Fig. 13(a) We then polished the piece oflutetium in air until the surface became shiny instead of\\\\nblack in order to remove the oxide from the surface.\\\\nLuH 3was synthesised by hydrogen absorption using\\\\nthe Sievert method. We used a HERA C2-3000 device to\\\\nmeasurethequantityofhydrogenabsorbed(ordesorbed)\\\\nby the piece of lutetium as a function of time. This is cal-\\\\nculated by measuring the hydrogen pressure variation in\\\\na sample holder of known volume. The measurement of\\\\nthe hydrogenation rate is performed out of equilibrium.\\\\nThe piece of polished lutetium (147.67mg) was placed in\\\\nthe sample-holder of the reaction chamber. The sample-\\\\nholder and compensation chambers were then pumped\\\\nfor one hour at ambient temperature to remove contam-\\\\ninating gases. The temperature was then increased to a\\\\nmaximum temperature of 500 °C at 10−5mbar and kept\\\\nstable for 4000s to outgas the container as much as pos-\\\\nsible. The temperature was then decreased to 200◦C,\\\\nand H 2gas at 4MPa was injected into the chamber. Af-\\\\nter 18hours, the weight percentage of absorbed H 2satu-\\\\nrated at 1.7 %which corresponds to the expected compo-\\\\nsition of LuH 3, as shown in Fig. 4 (though only the first\\\\n3.5hours are shown). After the synthesis, the sample-\\\\nholder was closed and transferred into an argon glove box\\\\nwhereitwasopenedtorecovertheLuH 3powder. Wecan\\\\nqualitatively compare the hydrogen concentration within\\\\nthe lattice to previous measurements by comparing the\\\\na-axis parameter [31, 73]. Previous work showed that a\\\\ngeneral trend amongst the trigonal/hexagonal rare-earth\\\\nhydrides is that the a-axis parameter decreases with in-\\\\ncreasing hydrogen concentration [73]. For our sample,\\\\na=6.173(1)Å whereas the a-axis parameter from Mans-\\\\nmann et al.wasdeterminedtobe6.163Å [31]. Similarly,\\\\nthea-axis value from Tkacz et al.is 6.50Å once con-\\\\nverted to the equivalent hexagonal structure [30]. There-\\\\nfore, the hydrogen concentration within our sample is\\\\nsimilar to previous samples and more densely packed\\\\nthan the sample of Tkacz et al.\\\\nAthinsampleofLuH 3waspreparedinadiamondanvil\\\\ncell (DAC) with culets of 800 µm diameter by pressing\\\\nthe synthesised powder between the two diamonds until\\\\nthe sample was approximately 5-10 µm thick. A stainless\\\\nsteel gasket was indented to a thickness of 80 µm and a\\\\nhole of 400 µm was drilled for the pressure chamber. A\\\\nrubysphereandasmallpieceofsiliconwereplacedinside\\\\nthe pressure chamber. Prior to loading the DAC, the\\\\nLuH 3sample was characterised by Raman spectroscopy\\\\nand X-ray diffraction (XRD) inside the unloaded DAC.\\\\nWe prepared three DACs in total with the trigonal\\\\nLuH 3powder. The first (DAC1) was largely discussed in\\\\nthe main text, and we used a gas loader (Top Industrie)\\\\nto load a mixture of nitrogen and helium. After purg-\\\\ning with helium, the system was filled with 10bar of N 2\\\\nand then 1500bar of helium. We estimate that the quan-\\\\ntity of N 2in the pressure chamber was 4nmol whilst the\\\\nquantity of LuH 3was 11nmol. The DAC was then sealed\\\\nat 0.1GPa and then we applied 1.9GPa and proceeded9\\\\nFIG. 4. The weight percentage of absorbed H 2by lutetium\\\\nas a function of time. After 3.5hours at 200◦C, 1.7 %of\\\\nabsorbed H 2is reached showing the successful synthesis of\\\\nLuH 3.\\\\nto characterise the sample by Raman spectroscopy and\\\\nXRD. The second DAC (DAC2) was loaded with pure\\\\nnitrogen at 1200bar, and the third DAC (DAC3) was\\\\ncryogenically loaded with pure nitrogen at 77K.\\\\nThe EDS measurements used a Bruker silicon drift de-\\\\ntector(SDD)mountedonaFESEMZEISSUltra+witha\\\\nworking distance of 8mm, a take-off angle (TOA) of 35 °,\\\\nand an acquisition time of about 2mins. To increase the\\\\nchance of observing nitrogen, which emits at 0.392keV,\\\\nWDS was performed with a JEOL-8800 Electron Probe\\\\nMicroAnalyzer (EPMA/Microsonde de Castaing). Qual-\\\\nitative analysis of nitrogen used a LDE1H synthetic su-\\\\nperlattice analyzing crystal (Si/W) and TAP for Lu-M α\\\\n(1.58keV). On the EPMA system, the TOA is 40 °. For\\\\nthe EDS, the electron-beam was rastered over an area of\\\\napproximately 2x2 µm2, whilst for the WDS a defocussed\\\\nspot of 10 µmwas used to limit the material degradation\\\\nby overheating or carbon contamination from the adhe-\\\\nsive tape. Both experiments used several voltages (from\\\\n5-15keV) though the ionisation efficiency of nitrogen is\\\\nenhanced at low voltages.\\\\nX-ray powder diffraction of the starting LuH 3was per-\\\\nformed immediately after the hydrogenation of lutetium\\\\nusing a D5000T diffractometer (Cu-K αradiation), at\\\\nambient pressure (and outside the DAC). The measure-\\\\nment was repeated several times (up to 9days after the\\\\nfirst measurement, and a final time after 6months) to\\\\ndetermine the effect of air exposure on LuH 3. The Ri-\\\\netveld refinements were done with FullProf software [74].\\\\nThe X-ray powder diffraction after loading at 1.9GPa in\\\\nDAC was performed on the ESRF beamline ID15B with\\\\nλ=0.411Å. Additional measurements on the LuH 3pow-\\\\nFIG. 5. Powder X-ray diffraction on the trigonal LuH 3. (a)\\\\nSample kept in glove-box and sealed between two pieces of\\\\ntape during the measurement. The reliability values for the\\\\nfit are (R p=7.25%, R wp=7.95%, R exp=1.79%, χ2=19.7). (b)\\\\nEven after 6 months of exposure to air, the quantity of Lu 2O3\\\\nimpurities did not change significantly with time. The pat-\\\\nterns are shifted by a constant offset to aid comparison.\\\\nder at ambient pressure were also performed on the same\\\\nESRF beamline. Calibration of the detector-to-sample\\\\ndistance, beam orientation, detector tilt with respect to\\\\nthe omega rotation axis, and the used wavelength was\\\\ndetermined by a Si powder standard (‘NIST 640 C’ from\\\\nNIST). The X-ray beam was focused to 4x3 µm2using\\\\nBe compound refractive lenses. 2D images were collected\\\\nwith a six degrees oscillation of the DAC using an Eiger\\\\n2X CdTe 9M photon counting detector from Dectris and\\\\nintegrated into a 1D pattern using the Dioptas software\\\\n[75]. Le Bail refinements (lattice parameter, peak pro-\\\\nfile, and background) on the loaded DAC at 1.9GPa were\\\\ndone using the GSAS-2 package [76].\\\\nPolarised Raman scattering was performed in quasi-\\\\nbackscattering geometry at 300K with an incident laser-\\\\nline at 532nm from a solid-state laser. The DAC was\\\\nplaced in a vacuum to avoid measuring the Raman re-\\\\nsponse of air. We used a laser power between 2.5-10mW\\\\nwith a typical spot size of 25 µm. The scattered light\\\\nwas analysed by a single grating and a triple grating10\\\\nsubtractivespectrometer, bothwereequippedwithliquid\\\\nnitrogen-cooled CCD detectors. The crossed and parallel\\\\npolarisation dependence was measured by changing the\\\\norientation of the polariser on the collection path. We\\\\nmeasured the Raman signal of pure LuH 3in the DAC\\\\nbefore and after loading the pressure medium.\\\\nS2: Trigonal lutetium trihydride\\\\nFig. 5(a) shows the pattern of the lutetium trihydride\\\\nimmediately after synthesis; it is well-described by a trig-\\\\nonal structure with some Lu 2O3impurities. After the\\\\nfirst XRD measurement, we left a small fraction of the\\\\nLuH 3powder exposed to air and measured the XRD sev-\\\\neral times over the course of 9days to check its stability.\\\\nThe rest of the powder was immediately stored under\\\\nvacuum or in an argon glove box. Figure 5(b) shows that\\\\ndespite being in contact with air, the Lu 2O3content is\\\\nsimilarwithintheerrorbar, i.e. 3.4(1) %vs3.2(1) %from\\\\nbefore. This remains true after 6months of exposure to\\\\nair.\\\\nFig 6.a shows the polarisation dependent Raman\\\\nspectra of the ambient pressure trigonal LuH 3below\\\\n955cm−1; at higher energies we do not identify any ex-\\\\ncitations that clearly originate from LuH 3. Within the\\\\naforementioned range, we observe 13 features (marked by\\\\narrows) which could account for most of the expected 17\\\\nphonons of the P3c1trigonal structure. Overall, we do\\\\nnot observe any significant differences between the differ-\\\\nentpolarisations. Theinsetshowsthelow-energyspectra\\\\ndown to 20 cm−1where we do not see any more notable\\\\nfeatures.\\\\nFig. 6(b) shows the Raman spectrum of trigonal LuH 3\\\\nbefore pressurising alongside the spectra of the translu-\\\\ncent part of the sample at high pressure and again at am-\\\\nbient pressure after releasing the pressure. Apart from\\\\na hardening of the phonon modes under pressure, we do\\\\nnot see any drastic change in the spectra. Importantly,\\\\nthe number of modes observed does not change over the\\\\npressure cycle, so it seems that this part of the sample\\\\nwas untransformed and largely unaffected by the high\\\\npressure. Why remains unclear.\\\\nDAC2 was primarily used to determine the pressure\\\\ndependence of the lattice volume. Initially, this was fol-\\\\nlowed at the ESRF Beamline up to 3GPa. Since the\\\\nsampleremainedtrigonal, itwasheatedto65 °Canddur-\\\\ning this process, the pressure increased up to 7GPa, yet\\\\nthe sample remained trigonal. The pressure was then in-\\\\ncreased further until 12GPa at room temperature with\\\\nXRD being measured at every pressure and the result\\\\nis shown in Fig. 7(a). The lattice parameters of the\\\\nrefined trigonal structure are shown in Fig. 7(b), whilst\\\\nFig. 7(c)showsthevolumedependenceonpressure. This\\\\nwas also calculated by Dangić et al.which is presented\\\\nalongside the volume dependence determined here and\\\\nFIG. 6. (a) Raman susceptibility of trigonal LuH 3at 300K\\\\nand 1bar measured in the unloaded DAC1 in cross and paral-\\\\nlel configurations. Arrows point to features of interest. Below\\\\n175cm−1, dataarescaledtooverlaywiththesingle-stagedata\\\\nmeasured at higher energies. The inset shows the unscaled\\\\ntriple-stagedataatlowenergy. Therawdataatambientpres-\\\\nsure from ref [8] are shown in grey and are scaled to aid com-\\\\nparison. (b) The pressure evolution of the translucent part of\\\\nthe sample at 300K in DAC1. The translucent part remained\\\\ntrigonal throughout the pressure cycle: from 0 to 1.9GPa and\\\\nback to 0GPa. Scaled comparisons with two other samples\\\\nin DAC2 and DAC3 (nitrogen pressure medium) at 2.0GPa\\\\nand 3.5GPa respectively are shown. Dotted lines show the\\\\nRaman shift of the dominant peak at ambient (red) and high\\\\npressure (blue and orange). Inset shows the pressure depen-\\\\ndence of the dominant peak and a linear fit.\\\\nshows a similar trend with a small offset. After that, the\\\\npressure was decreased to 2GPa, whereupon the Raman\\\\nspectroscopy was measured which is presented in figure\\\\n6(b) Throughout all of the pressure changes the sample\\\\nremained trigonal.11\\\\nFIG. 7. (a) X-ray diffraction patterns of the trigonal LuH 3\\\\nphase under pressure in DAC2. (b) Variation of the fitted\\\\nlattice parameters with pressure for the trigonal phase. (c)\\\\nThe lattice volume versus pressure data for the trigonal phase\\\\nwith a Birch-Murnaghan fit. Predictions by Dangić et al.[32]\\\\nare also shown for comparison.Conditions λ(Å)a-axis (Å) c-axis (Å)\\\\nBefore (1bar/300K) 1.546.1680(8) 6.422(1)\\\\n3.5GPa before heating 0.566.111(5) 6.335(9)\\\\n3.5GPa after heating 0.566.113(6) 6.338(9)\\\\nAfter decompression 0.566.1744(4) 6.421(8)\\\\nTABLE II. Refined attice parameters of the LuH 3sam-\\\\nple loaded with cryogenic liquid nitrogen (DAC3) at several\\\\nstages throughout the synthesis process.\\\\nAfter cryogenically loading DAC3 and warming to\\\\nroom temperature, the pressure was determined to be\\\\n3.5GPa. At this pressure, both the Raman and XRD\\\\nconfirmed that the structure remained trigonal (see Figs.\\\\n6(b)and8respectively). HerealaboratoryK α-Agsource\\\\n(λ=0.56Å) was also used to measure the XRD. The\\\\nDAC was then heated at 65 °C for 24h as was done for\\\\nboth the sample in the main text and the Dasenbrock-\\\\nGammon sample [8]; the resulting XRD pattern is shown\\\\nin Fig. 8 and there is no measurable difference within the\\\\nerror, as shown by the refined lattice parameters in ta-\\\\nble II. Overall we do not reproduce the cubic structural\\\\ntransition in this cell either. Upon decompression, the\\\\nrecovered sample remained trigonal but with a slightly\\\\nlarger a-axis than the original sample before compres-\\\\nsion, though this could be intrinsic hysteretic behaviour\\\\nof the sample caused by compression and decompression.\\\\nFIG.8. PowderX-raydiffractionofthetrigonalLuH 3incryo-\\\\ngenically loaded nitrogen pressure medium (DAC3). Black\\\\nand red lines show the data at 3.5GPa before and after heat-\\\\ning respectively; they are effectively identical and overlay al-\\\\nmost perfectly. Blue data show the pattern after releasing the\\\\npressure.\\\\nIn both cells loaded with pure nitrogen (DAC2 and\\\\nDAC3), weobserveRamanspectrathatresembletrigonal\\\\nLuH 3at high pressure, as shown by Fig. 6. These trigo-\\\\nnal samples and the small trigonal part in DAC1 all show\\\\naverysimilarlinearhardeningwithpressureforthedom-\\\\ninant phonon mode, as shown by the inset of Fig. 6(b).12\\\\nSato et al.showed that the Raman spectra of pressurised\\\\nSiO 2glass change when in a helium pressure medium, as\\\\nthe helium atoms occupy interstitials within the silicate\\\\nnetwork [40]. Here, we do not observe any significant dif-\\\\nference between the trigonal LuH 3samples loaded in the\\\\npressure media, and the hardening of the phonons under\\\\npressure follows the same behaviour in all of the pressure\\\\nmedia. This leads us to believe that the helium pressure\\\\nmedium is not causing the structural change in DAC1.\\\\nConsidering the effects of the pressure media themselves,\\\\nsince both helium and nitrogen perform well as hydro-\\\\nstatic pressure media to at least 10GPa [77], we do not\\\\nexpect significant uniaxial effects below 2GPa. So the\\\\ndifference in hydrostaticity is unlikely to be the origin of\\\\nthe difference between DAC1 (with transformation) and,\\\\nDAC2 and DAC3 (without transformation).\\\\nS3: Transformation of the LuH 3sample\\\\nFigs. 9(a)and9(b)showwide-rangeRamanspectraon\\\\ntheambientpressuretrigonalLuH 3andthehigh-pressure\\\\ncompoundofDAC1. Herethemodesinthehigh-pressure\\\\nstructure clearly do not resemble the modes in the trig-\\\\nonal structure. The spectra of the high-pressure phase\\\\nfor multiple spots on the sample also show the same fea-\\\\ntures, thoughthebackgrounddoeschange. Thelocations\\\\nof these different spots are shown in the inset image. In\\\\ntable III, we write the energies of the excitations seen\\\\nin the original trigonal LuH 3and the high-pressure com-\\\\npound.\\\\nCompounds LuH 3High pressure compound\\\\n(0 GPa) (1.9 GPa)\\\\nEnergy 100.2 128\\\\n(cm−1) 109.4 164\\\\n117.4 202\\\\n132.6 260\\\\n147.5 1141\\\\n368.4 1241\\\\n416.8\\\\n454.2\\\\n550.2\\\\n702.2\\\\n755\\\\n829\\\\n861.8\\\\n1039\\\\nTABLE III. Raman modes energy measured on trigonal LuH 3\\\\nat ambient pressure and the high-pressure compound mea-\\\\nsured at 1.9GPa in DAC1. In italics, are the modes which\\\\nare difficult to identify.\\\\nTo complete the synthesis as described by Dasenbrock-\\\\nGammon et al, DAC1 was heated at 65◦C for 24h at\\\\n1.9GPa. Fig. 9(c) shows the resulting Raman spectra;\\\\nnot much has changed.\\\\nWe then opened the DAC1 in helium gas (20bar) to\\\\navoid contact with air. Then we slightly closed the DAC\\\\nFIG. 9. (a) Raman susceptibility of the trigonal LuH 3at am-\\\\nbient pressure. (b) Raman susceptibility of the compound at\\\\n1.9GPa (DAC1). Data on three different spots are presented,\\\\nand the inset shows the locations of the spots on the sample.\\\\nBelow 175 cm−1, triple-stage data are overlaid on the high-\\\\nenergy spectra. (c) Raman susceptibility of the high-pressure\\\\nphase before and after annealing at 1.9GPa and 65◦C for\\\\n24h. The purple data were scaled such that the intensity of\\\\nthe high-energy mode is similar. (d) through to (f) show the\\\\nRaman susceptibility of the annealed sample at several times\\\\nafter opening the DAC. (g) The raw Raman spectra of part\\\\nA of the sample from Dasenbrock-Gammon et al.at ambient\\\\npressure and at 2.17GPa are presented [8].13\\\\nFIG. 10. Polynomial background subtracted Raman spectra\\\\nshowing the time evolution of the sample at 0GPa after pres-\\\\nsurising to 1.9GPa and heating to 65 °C. We also overlay data\\\\nfrom Dasenbrock-Gammon et al.[8] and Jiang et al.[34] to\\\\ncompare against ‘part B’ of their sample and Lu 2O3, respec-\\\\ntively.\\\\nto keep the sample in a helium environment and remea-\\\\nsured the sample at essentially ambient pressure. The\\\\nresults are shown in Figs. 9(d) to 9(f). Shortly after\\\\nopening, the spectrum resembles the cubic phase with\\\\na peak located just below 250 cm−1and what could be\\\\na broad and weak remainder of the peak at 1240 cm−1.\\\\nHowever after one day, this high-energy peak has disap-\\\\npeared but the low-energy peak remains. Fig. 9(f) shows\\\\nthe spectrum after several days (during which time the\\\\nsample was stored under vacuum), and clearly the struc-Fm3m-type Ia¯3-type P63/mmc LuH x\\\\na-axis (Å) a-axis (Å) a-axis (Å) c-axis (Å)\\\\n1 4.798 10.427 3.529 5.588\\\\n2 4.806 10.433 - -\\\\n3 4.776 - 3.515 5.589\\\\n4 4.773 - 3.5099 5.584\\\\n5 4.796 10.402 - -\\\\n6 4.785 10.409 3.527 5.561\\\\n7 4.781 10.399 - -\\\\n8 4.788 10.410 3.524 5.583\\\\nAverage 4.79(1) 10.41(1) 3.521(7) 5.58(1)\\\\nTABLE IV. Lattice parameters from Le Bail refinements of\\\\nthe three phases in the sample from DAC1 released at am-\\\\nbient pressure and measured in several different locations on\\\\nthe sample. A hyphen means that the given phase was not\\\\nobserved in that location.\\\\nture has changed once again. This spectrum resembles\\\\nneither the cubic nor the trigonal phase. In Fig. 10, we\\\\ncompare the background-subtracted signals of the data\\\\nin Figs. 9(d)-(f) against the spectra of Lu 2O3[34] and\\\\n‘part B’ from Dasenbrock-Gammon et al.[8]. There is\\\\nno strong resemblance between either of the other com-\\\\npounds, with the exception of the most intense peak of\\\\nLu2O3, which would have to be considerably broadened,\\\\nand the low-energy peaks of ‘part B’, but the rest of the\\\\nspectrum is different.\\\\nSubsequently, we measured the XRD after releasing\\\\nthe pressure, and the corresponding diffraction XRD im-\\\\nage is shown in Fig. 11(b), whereas Fig. 11(a) shows the\\\\nhigh-pressure phase. The most evident change is that the\\\\n0GPa XRD image has become spotty instead of form-\\\\ning continuous rings. This shows that the crystalline do-\\\\nmain sizes are larger than the X-ray beam size (4x3 µm2)\\\\nwhich means that we can no longer fit the patterns with\\\\nRietveld refinements. Qualitatively, in the ambient pres-\\\\nsure patterns, we see three phases as shown in Fig. 12.\\\\nWe measured 8 different spots. Firstly, we observe sim-\\\\nilarFm3mandIa¯3-type structures to those measured\\\\nat high pressure, but in addition we observe a P63/mmc\\\\nphase. Fm3mphase is present in every measured spot,\\\\nbut this forms either a biphasic mixture with the Ia¯3-\\\\ntype (3/8 spots) or the hexagonal phase (2/8 spots), or it\\\\nforms a triphasic mixture (3/8 spots). The refined lattice\\\\nparameters of the measured phases in different locations\\\\nare shown in table IV.\\\\nTo understand this, we must first consider the bi-\\\\nnary mixture phase diagram of lutetium and hydrogen\\\\n[57, 58, 60, 78]. For low hydrogen concentrations up to\\\\n0.2H/Lu, a pure hexagonal P63/mmcLuH δ(0≤δ≤\\\\n0.2) forms; the lattice parameters of which increase with\\\\nincreasing hydrogen concentration until they saturate at\\\\na=3.5240Å and c=5.6050Å for LuH 0.2[78]. Both of\\\\nour average values of a=3.521(7)Å and c=5.58(1)Å in-\\\\ndicate a lower hydrogen concentration: the values of a\\\\nandcimply δ=0.16(7) and δ=0.09(3), respectively. Be-14\\\\nFIG. 11. 2D XRD images of the sample after heating to\\\\n65°C (a) at 300K and 1.9GPa and (b) after the pressure was\\\\nreleased. Temporally, (b) was measured between the Raman\\\\ndata displayed in Figs. 10(e) and 10(f), i.e. between 1day\\\\nand 5days after opening the cell. Both XRD images were\\\\nobtained with the sample in the DAC.\\\\nyond 0.2H/Lu, a binary mixture of the P63/mmcLuH δ\\\\nand an Fm3mphase forms. There is uncertainty where\\\\nthe end of this binary mixture ends: some sources say\\\\n≈0.6H/Lu [57, 58] while another says 1.8H/Lu [60].\\\\nThe latter concentration forms a compound that is ap-\\\\nproximately the same as LuH 2which has a lattice pa-\\\\nrameter of a=5.035 Å [78]. This value is much larger\\\\nthan our average value of 4.79(1)Å. But in the instance\\\\nthat 0.6H/Lu is the beginning of the binary mixture, it\\\\nis then probable that the low concentration Fm3mphase\\\\nwould have a much smaller lattice parameter than LuH 2\\\\nwhich could then be close to our value. Finally and as\\\\ndiscussed in the main text, the lattice parameter of the\\\\nIa¯3-type structure expands when the pressure is released\\\\nand becomes larger than the ambient pressure value of\\\\n10.38Å for Lu 2O3, therefore we conclude that the Ia¯3-\\\\ntype phase is a distinct compound from Lu 2O3.Here and in the main text, we consider the decompo-\\\\nsition of the initial LuH 3into lower hydrides. This must\\\\nresult in the formation of H 2which should in principle\\\\nbe measurable by Raman spectroscopy. At high energy,\\\\nthere exists a well-known hydrogen vibron excitation at\\\\napproximately 4200 cm−1at low pressure and 300K [79–\\\\n81]. However, this vibron is inherently weak and gener-\\\\nally only visible with a pure hydrogen pressure medium\\\\nor with ammonia borane after laser heating due to the\\\\nlarge concentration of hydrogen present. In our work, the\\\\nproposed decomposition of LuH 3to LuH 2+xwould only\\\\nproduce a fraction of a hydrogen atom per unit cell and\\\\nthereforealowconcentrationofhydrogen; thustheinten-\\\\nsityofthevibronwillbeweaker. Furthermore,thehydro-\\\\ngencanescapethepressurecellwhichfurtherreducesthe\\\\nquantity present and diminishes the intensity. As a result\\\\nof all of these reasons, we did not observe the high-energy\\\\nhydrogenvibron. Therealsoexistsaweakerhydrogenex-\\\\ncitationatapproximately1044 cm−1[81], whichisclearly\\\\nobservable in the data of Dasenbrock-Gammon et al.in\\\\nFig. 3(c) of the main text. This is due to their use of\\\\na hydrogen pressure medium, but despite that, the exci-\\\\ntation remains weak. Since we did not use a hydrogen\\\\npressure medium and the aforementioned reasons, it is\\\\nnot surprising that we do not observe it.\\\\nS4: EDS and WDS analysis of the recovered sample\\\\nScanning electron microscopy with X-ray energy dis-\\\\npersive spectroscopy (EDS) and wavelength dispersive\\\\nspectroscopy (WDS) were used to analyse the compo-\\\\nsition of the pure lutetium and recovered sample. Fig.\\\\n13(a) shows the EDS spectra of the recovered sample af-\\\\nter pressuring at 1.9GPa and heating at 65 °C, and pure\\\\nLu after polishing; all spectra were normalised by the\\\\nmaximum intensity of a given spectrum. At high ac-\\\\ncelerating voltages, one preferentially excites the energy\\\\nlevels of the heavier atoms, whilst at low voltages, the\\\\nsignal of lighter elements becomes more intense. This\\\\nis most clearly seen in the intensity of the O-K αpeak\\\\nwhich grows in intensity relative to the Lu-M αpeak at\\\\nlow voltages. Thus to find nitrogen, lower accelerating\\\\nvoltages should be used.\\\\nFirstly though we should comment on the other atoms\\\\ndetected: oxygen and carbon. As mentioned before, oxy-\\\\ngen originates from Lu 2O3and is also present in freshly\\\\npolished lutetium metal. Its presence is not a surprise.\\\\nThe carbon originates from the tape used to attach the\\\\nsampletothesampleholder,asaconductivesurfacemust\\\\nbe used, therefore this is also expected.\\\\nThe characteristic K αemission energy of nitrogen is\\\\nsituated at 0.392keV as indicated in fig 13.a. However,\\\\nwithin the noise of the measurement for these EDS mea-\\\\nsurements, there is no indication of nitrogen in the struc-\\\\nture. We also note that there is very little difference be-15\\\\nFIG.12. XRDpatternsonthreedifferentspotsofthereleased\\\\nsample that were measured at the ESRF with λ=0.411Å. We\\\\nidentify three different phases: Fm3m-type, Ia¯3-type, and\\\\nP63/mmcLuH xwhich are all fitted with Le Bail fits.\\\\ntween the recovered sample and the pure lutetium. We\\\\nalso used WDS which has superior resolving power, as\\\\nFIG. 13. (a) EDS measurements on the high-pressure sam-\\\\nple of DAC1 after releasing the pressure and polished pure\\\\nlutetium at different accelerating voltages. Several key emis-\\\\nsion lines are indicated. (b) and (c) A comparison between\\\\nthe 5keV measurements with EDS and WDS on the recovered\\\\nsample. The WDS measurements were scaled on the O-K α\\\\nand Lu-M αlines respectively to aid comparison.\\\\nshown in Figs. 13(b) and 13(c) by the narrower O-K α\\\\nline and the Lu-M α+Lu-M βline being clearly resolved\\\\ninto the two spectral lines. This helps to distinguish the\\\\npotential nitrogen excitation from the nearby carbon and\\\\noxygen excitations. With the WDS measurements, we\\\\nused the same low voltage as for the EDS such that we\\\\ncould preferentially observe nitrogen, but there is no ob-\\\\nservable feature at the N-K αexcitation energy, "},"Similar Full Paper":{"kind":"string","value":"{'Effect of CoO loading on electrochemical properties of activated carbon from sugarcane bagasse': 'Title: Effect of CoO loading on electrochemical properties of activated carbon from sugarcane bagasse\\\\nJournal of Analytical and Applied Pyrolysis 168 (2022) 105724Available online 23 September 20220165-2370/© 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).Effect of heating rate and H3PO4 as catalyst on the pyrolysis of agricultural residues Behnam Hosseinzaei a, Mohammad Jafar Hadianfard a,*, Ramiro Ruiz-Rosas b, Juana M. Rosas b, José Rodríguez-Mirasol b,*, Tomás Cordero b a Department of Materials Science and Engineering, School of Engineering, Shiraz University, Shiraz, Iran b Universidad de Málaga, Andalucia Tech., Departamento de Ingeniería química, Campus de Teatinos s/n, 29010 Málaga, Spain A R T I C L E I N F O Keywords: Agricultural residue Acid treated biomass Fast pyrolysis Porosity Bio-oil Syngas A B S T R A C T This study reports the effect of heating rate and the addition of H3PO4 on the pyrolysis of three representative agricultural wastes of different lignocellulosic composition, namely pistachio shell, bitter orange peel, and saffron petal. Pyrolysis was carried out at 500 ◦C in a fixed-bed, lab scale reactor. Slow pyrolysis provided lower water contents in the liquid fraction. Fast pyrolysis increased the liquid yield for all the feedstocks, promoting the formation of phenolic, ketone/aldehyde compounds. It also enhanced the formation of water for all the agri-cultural residues. In addition, the energy content in the gas fraction is promoted due to a higher concentration of light hydrocarbons, methane, and hydrogen. However, when high inorganic matter is found in the feedstocks, the formation of CO2 is favored, hindering the energy improvement. The treatment of the biomass with H3PO4 significantly increased the solid fraction, producing a huge porosity development in the char (surface area over 1600 m2/g in pistachio shell product), at the cost of liquid fraction, which is mostly composed of water, with small amounts of acetic acid, phenol and toluene. The results pointed out that pyrolysis of agricultural waste can be targeted to achieve different products by switching pyrolysis conditions such as the heating rate and the treatment of the biomass with H3PO4. 1. Introduction The increasing world population and development brings higher energy demand, which leads to overconsumption of nonrenewable fossil sources. Exploiting these fossil fuels at high pace depletes their reserves and emits toxic gases (COx, NOx, SOx), which are harmful to humans and their environment [1,2]. Biomass is an abundant and cost-effective renewable energy, which is mostly found in every world region and can help in replacing the use of fossil fuels [3]. Agricultural residues, as lignocellulosic biomass, can be considered as an emerging source of energy and chemicals. In this study, three agricultural residues with different origin and composition, such as pistachios shell, bitter orange peel, and saffron petals are chosen to be used as feedstocks for their valorization into high value-added products. In the case of pistachio shell, the countries with the highest production of pistachios in 2019 were Iran (571,000 tons), the United States (484,000 tons), and Turkey (267,000 tons); their combined share was 88% of the total production of the world [4]. On average, the empty shell wastes, including the hard shell, form around 15% of the total product. As an example, an amount of 77,550 tons of waste including hard shells is formed only in Iran per year [5]. Citrus fruits are one of the most important agricultural products in the world. In Spain, nearly 6.5 million tons of citrus are produced per year, of which over 1 million tons waste are left annually [6]. Orange industry generates different residues in the form of seeds, pulp, and peel. Specifically, 20% of the orange is orange peel, so only in 2018 were more than 15 Mt of orange peel generated in the world [7]. Saffron is another agricultural product, whose industry produces high amounts of residues. Saffron production reached more than 350 tons in 2019 [8] and annually 194,445 tons of petals are wasted as part of the production process in Iran, which accounts for 90% of the world’s cultivation area. The most common biomass thermochemical conversion processes are gasification, combustion, and pyrolysis. During the latter process, biomass is thermally decomposed in an oxygen free environment to gas, liquid, and solid fractions [9–11]. Depending on temperature, heating rate, and residence time, the * Corresponding authors. E-mail addresses: hadianfa@shirazu.ac.ir (M.J. Hadianfard), mirasol@uma.es (J. Rodríguez-Mirasol). Contents lists available at ScienceDirect Journal of Analytical and Applied Pyrolysis journal homepage: www.elsevier.com/locate/jaap https://doi.org/10.1016/j.jaap.2022.105724 Received 25 July 2022; Received in revised form 19 September 2022; Accepted 21 September 2022 Journal of Analytical and Applied Pyrolysis 168 (2022) 1057242pyrolysis process can be grouped into three types: slow, fast, and flash pyrolysis. Slow pyrolysis is implemented at low heating rates and high residence times, within a temperature range of 400–700 ◦C, meanwhile fast pyrolysis occurs under higher heating rates, at temperatures around 500 ◦C, and low residence time of evolved gases. Flash pyrolysis is aimed to heat the biomass with a very high heating rate at a temperature range of 450–900 ◦C [12–14]. Fast pyrolysis is considered as one of the promising thermochemical technologies that provide sustainability in many aspects like energy, economy, environment, and well-being of society. This process has been performed on different biomass types such as wood, agricultural resi-dues, and domestic industrial wastes [15]. During the fast pyrolysis process, biomass is decomposed to vapor/gas and solid residue. The vapor is quickly cooled to room temperature for avoiding secondary cracking reactions, and therefore leads to an increase of the liquid fraction, which contains different organic compounds with a wide range of molecular weights and its composition and quality are heavily dependent on the composition of biomass feedstock. Bio-oil is an in-termediate product, which can be easily stored and transported to be processed for fuel and chemicals production [16]. Even though fast pyrolysis is thoroughly studied in the literature, the systematic comparison between slow and fast pyrolysis of varied agri-cultural residues is not deeply analyzed, with only a few examples being found in the literature. In this context, Taib et al. investigated the fast pyrolysis process to produce the bio-oil from of banana pseudo-stem, which at optimum conditions (T = 500 ◦C and residence time=1.02 s) total liquid yield of 39.4 wt% was obtained [17], however, the distri-bution of products of slow pyrolysis is not reported. In another study, in the slow and fast pyrolysis of Cherry seeds (CWS) and cherry seeds shells (CSS), maximum bio-oil yields were about 44 wt% at 500 ◦C for both biomasses, whereas the bio-oil yields obtained under slow pyrolysis for SWS and CSS were 21 and 15 wt%, respectively; a fluidized bed reactor was used for fast pyrolysis experiments, while slow pyrolysis was stud-ied in a fixed bed reactor [18]. Yang et al. also investigated the fast and slow pyrolysis of different parts of eastern redcedar under fast and slow pyrolysis at 450 and 500 ◦C; a pyrolysis probe attached to GC-MS is used to simulate fast pyrolysis, while slow pyrolysis is performed in a high-pressure batch reactor [19]. On the other hand, pretreatment of biomass is sometimes performed before pyrolysis in order to extract value-added compounds, modify the density or adapt the composition for achieving optimal results. For instance, treatment of the raw materials with acidic and alkaline agents affects the structure of the biomass, which changes the product yields distribution and properties of each pyrolysis product [20–22]. Various chemical agents, such as H2SO4, NaOH, KOH, ZnCl2, and H3PO4, have been used for this aim [23]. One of the most striking features of chemical treatment is producing solid, known as activated carbon, with high surface area and pore volume as well as the presence of surface func-tional groups. For example, activated carbon from different agricultural wastes such as pistachio shell [23], date pits [24], jackfruit peel [25], and orange skin [26] has been already produced by chemical activation of different agricultural wastes such as pistachio shell [23], date pits [24], jackfruit peel [25], and orange skin [26]. Since most efforts are devoted to study the properties of the pyrolysis solid product, little is known about the composition and yields of liquids and gas phases ob-tained as coproducts in these treatments [27,28]. The objective of this study was to determine the potential benefits of fast and H3PO4-treated pyrolysis to improve the production of bioenergy and products with higher added value in the valorization of residual biomasses. Bearing this aim in mind, the slow, fast, and catalyzed py-rolysis with H3PO4 are performed using a similar fixed-bed reactor configuration, so that meaningful comparison between processes can be drawn. Such a study could bring light into which lignocellulosic feed-stocks are more adequate for the production of the different pyrolysis products, when moving from slow to fast pyrolysis, or even catalyzed slow pyrolysis. Considering the importance and great volume of residues produced from pistachio, bitter orange and saffron, representative raw material of three major families of agricultural residues: nutshells, fruit skin and petals, as well as their different lignocellulosic composition, these wastes were chosen as models of three different families of agri-cultural residues (shells, peels, and petals) for validating the conclusions for a wide range of raw materials. For obtaining a good picture of the effects of these variables on the pyrolysis process, attention to the full characterization of the whole products is needed. Furthermore, in order to evaluate their potential as energy products, the yields and heating values of the different fractions have been also determined. In addition, the composition and porosity of the solid fractions have been also characterized for assessing their potential use as adsorbents. Apart from these, the distribution of pyrolysis products of saffron petals under different conditions is reported for the first time. 2. Materials and methods 2.1. Material preparation Raw materials were Pistachio shell (PS), bitter orange peel (OP), and saffron petal (SP). PS and SP were provided from Khorasan Razavi, northeast of Iran, and OP was gathered from Fars province, south of Iran. Prior to their use, the feedstocks were dried at 105 ◦C in an air-dry oven for 12 h to remove the moisture. The dried samples were then milled and sieved to a particle size between 300 and 500. The prepared samples were stored in plastic flasks for their future uses. The physicochemical of the biomasses are reported in Table 1. More information can be found in our previous work [29]. Regarding SP biomass, to our knowledge, there is no published work on the content of the biopolymer components in the literature, except a study conducted by Fahim et al., who reported the results based on the dry weight (% w/w) as 10.2 protein, 8.8 fiber, 7.0 ash, 5.3 fat [30]. The acid-treated process was as follows: The prepared samples from the previous step were impregnated by incipient wetness with 85% (w/ w) H3PO4 aqueous solution at room temperature and dried overnight at 60 ◦C in an oven. The impregnation ratio, R, (H3PO4/ precursor mass ratio) was set to 3 in accordance with a previous study regarding the preparation of H3PO4-activated carbons from orange peel [33]. 2.2. Ultimate and proximate analysis of the raw materials The elemental analysis was verified by a TruSpec micro CHNSO (Leco) analyzer to determine the mass fractions of carbon, hydrogen, nitrogen, and sulfur. The proximate analysis of biomasses was per-formed by a thermogravimetric analyzer (Q500, TA Instruments, USA) [34]. In the thermogravimetric analysis (TGA) experiment, 10 mg from each sample was loaded into the platinum container, and then heated from ambient temperature to 900 ◦C at a heating rate of 10 ◦C/min under a flow rate of 60 mL/min of nitrogen as carrier gas. At that point, carrier gas is switched to air and the temperature is held until constant weight is obtained. 2.3. Experimental setup 2.3.1. Slow and H3PO4-treated biomass pyrolysis experiments The experiments for slow and H3PO4- treated biomass pyrolysis were performed under nitrogen atmosphere (flow rate: 150 mL/min STP) using a vertical tubular reactor (diameter: 2 cm). The reactor was loaded with a bed of 2–3 g of the selected feedstock, which was hold in the isothermal zone of the reactor using a compacted quartz wool layer. The experiments were carried out under atmospheric pressure, at 500 ◦C and at a heating rate of 10 ◦C/min. The bed temperature was tracked using an internal thermocouple. After pyrolysis, the furnace was cooled down to room temperature. The solid residue was recovered from the reactor at the end of the experiment, whereas the bio-oil was obtained at the reactor outlet from a condenser at ca. 0 ºC (mixture of acetone and ice is B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057243used as coolant). The experiments were repeated at least three times for each biomass in each process. The mass of liquid product and solid residue (char) were weighted, and the gas yield was confirmed by difference. 2.3.2. Fast pyrolysis experiment Fig. 1 shows the schematic diagram of the system used in the fast pyrolysis process. The same reactor described in the previous section is used in this system. On the top of the reactor, biomass was loaded inside a dropper consisting of a 4 cm height chamber with 25 mm of internal diameter, gas inlet and outlet for purging of the chamber, and two ball valves for dropping biomass inside the heated reactor and loading new biomass into the dropper. Inside the reactor, a biomass holder zone consisting of a quartz wool layer was placed in the isothermal region so that biomass can be dropped into the previously heated reactor. An additional layer of 5 g of silicon carbide was used as an inert heat carrier. The reactor was mounted inside an electrical vertical furnace and was insulated to minimize heat loss. The experiment was performed under nitrogen atmosphere (flow rate: 150 mL/min STP) and the bed tem-perature was tracked using an internal thermocouple. The effect of using different biomass loadings on the reactor temperature was also studied. A thermocouple attached to a data logger was placed on top of the SiC layer for this purpose. The temperature of the bed was 495 ºC after one second when 100 mg of biomass were dropped into the reactor. Under these conditions, the heating rate was estimated to be higher than 100 ºC/s. In accordance with this finding, the fast pyrolysis experiments consisted of 8–10 sequential drops of 100 mg of the chosen biomass. The interval between the different loadings was 2–3 min, time necessary to obtain a negligible CO and CO2 evolution. While the biomass of the previous loading was being pyrolyzed, a new biomass sample was loaded in the dropper through the top valve and purged with nitrogen before being introduced into the reactor. Regarding the recovery of the products, the reactor was connected to two cooled condensers to gather the bio-oil. The biochar was collected from the reactor after the exper-iment was finished. 2.4. Analysis of products 2.4.1. Analysis of bio-oil composition by GC/MS Bio-oil was recovered from the condenser at the outlet of the py-rolysis reactors. The bio-oil was dissolved in acetone (analytical grade, Sigma-Aldrich) in 1:100 vol ratio. All the resulting mixtures were analyzed by Gas Chromatography–Mass Spectrometry (GC/MS), using an 7000D GC/MS Triple Quad (Agilent Technologies, USA), equipped with an Agilent DB-624 column (60 m × 0.250 mm×1.4 µm) and HP- 5 ms column (30 m × 0.250 mm×0.25 µm), FID and mass spectrometer detector (MS). Ions were detected in full scan mode (mass range from 15 to 400 m/z), with an electronic impact of 70 eV. Identification of com-pounds was achieved by comparing the mass spectra with the NIST MS Search 2.0 mass spectral library. The water content in the bio-oil was Table 1 Physicochemical composition of different lignocellulosic biomasses. Proximate analysis (wt%) Ultimate analysisa (wt%) biochemical composition [31,32] M VM Ash FC C H N S Ob Ce HC L P EX PS 2.3 78.1 0.4 19.2 48.1 6.4 0.1 < 0.1 45.4 51.2 21.5 21.5 – 5.8 OP 1.0 67.5 4.9 26.6 44.4 6.2 1.0 0.1 48.3 25.1 10.2 4.3 34.0 26.4 SP 2.5 65.8 6.4 25.3 49.1 6.5 2.0 2.0 40.4 – – – – – Composition of ash (wt%) K Ca Mg P Si other PS 0.1 – – – – 0.2 OP 1.9 2.1 0.4 0.1 – 0.3 SP 3.0 0.6 0.5 0.7 0.5 0.9 M= moisture, VM= volatile material, A=ash, FC=fixed carbon. Ce=cellulose, HC=hemicellulose, L=lignin, P = pectin, EX=extractive. a dry-ash free (daf) basis. b was calculated by difference. O = 100 – C – H – N – S. Fig. 1. Schematic of the fast pyrolysis set up (1) nitrogen cylinder (2) mass flow controller (3) temperature controller (4) thermocouple (5) N2 inlet (6) ball valve (7) biomass dropper (8) reactor (9) pyrolysis furnace (10) quartz wool (11) first condenser (12) second condenser (13) three-way gas collecting (14) NDIR sensor (15) computer. B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057244measured by Karl-Fisher titration (KF V20, Mettler Toledo). Bio-oil acidity was determined using a pHmeter (HI8424, Hannah). 2.4.2. Analysis of the non-condensable gases Analysis of the gas fraction from slow and H3PO4-treated biomass pyrolysis experiments was performed at the condenser outlet. The gas composition was determined with a Perkin Elmer Auto system GC equipped with a packed column (Hayasep-D 100–120 mesh, PE) with FID and TCD detectors. Identification and absolute quantification were carried out by external calibration with a standard commercial mixture of gases obtained from Linde. The gases evolution from the fast pyrolysis was analyzed with two analyzers. CO and CO2 concentrations were determined online by a non- dispersive infrared (NDIR) gas analyzer (Siemens Ultramat 23). Hydrogen, methane, and light hydrocarbons (C2-C3) were determined offline by injecting the gas stored in the gasbag into a Perkin Elmer Auto system GC equipped with a packed column (Hayasep-D 100–120 mesh, PE) with TCD and FID detectors. The LHV values have been determined following the equation deduced by Lv et al. [35]. 2.4.3. Analysis of the solid fraction In order to analyze the solid fraction obtained from the acid treated samples, the excess of phosphoric acid was removed by washing with distilled water at 60 ◦C until neutral pH and negative phosphate analysis in the eluate. The respective solids obtained are denoted as PSC, OPC and SPC. A truspec micro CHNSO analyzer, from Leco was used to determine the mass fractions of carbon, hydrogen, nitrogen, and sulfur for the elemental analysis. Proximate analysis was carried out to calculate the moisture contents, fixed carbon, volatile matter, and ash contents. This analysis was conducted by using a thermogravimetric analyzer (TGA/ DSC1, from Mettler Toledo). The higher heating value (HHV) of the solids was calculated based on the following equation derived from Cordero et al. [36]. HHV = 0.3543 × FC+ 0.1708 × VM (MJkg, dry basis) (2) where VM stands for volatile matter (%) and FC is fixed carbon (%) in the solid. The porous structure of the solid fractions was characterized by N2 adsorption–desorption at − 196 ◦C and by CO2 adsorption at ◦C. The measurements were performed in an ASAP 2020 model equipment of Micromeritics. Prior to the experiments, the samples were degassed overnight at 150 ◦C. From the N2 adsorption isotherm, the apparent surface area (SBET) was calculated by applying the BET equation. Con-sistency criteria as suggested by Rouquerol was applied to ensure that BET values are properly reported [37]. Application of the Dubinin-Radushkevich method to the N2 and CO2 adsorption isotherm provided the micropore volume (VDR N2 and VDR CO2). Finally, the mesopore volume (Vmes) was calculated as the difference between the adsorbed volume at a relative pressure of 0.96 and the micropore vol-ume (VDR N2). 3. Results and discussion 3.1. Yield of products Fig. 2 shows the yield of the corresponding products for the different type of pyrolysis for PS, OP, and SP. Fast pyrolysis shows higher liquid yields compared to those obtained by slow pyrolysis. The high bio-oil yields derived from fast pyrolysis are due to its high heating rate and short gas residence time, which enhance the rapid fragmentation of biomass and mitigate the secondary cracking of tar [18,38]. Higher bio-oil fractions under fast pyrolysis were also reported by Pütün et al., which studied the slow and rapid pyrolysis of pistachio shell in a tubular reactor. Temperature, heating rate and nitrogen flow were variable parameters in their work. The highest bio-oil yield was 27.7% when temperature, heating rate and nitrogen flow were 500 ◦C, 300 ◦C/min, and 100 cm3/min, respectively [38]. The superior yield obtained at the same temperature (almost double, 53 wt%) is probably related to the shorter residence time of gas in this system (aprox. 2 s for the isothermal region in the system herein used), which avoids secondary cracking reactions. In the case of bitter orange peel, no direct comparison using a similar system can be found in the literature. Alvarez et al. reported bio-oil yields of ca. 55% wt and char yields of 27–33% for fast pyrolysis of citrus waste using a conical spouted bed reactor [39], whereas Aguiar et al. studied slow pyrolysis of orange peel in fixed bed pyrolysis at different temperatures, obtaining char yield of 28–29% and liquid yields of ca. 40%, with a large content of water, at 450 ºC [40]. The larger liquid yield obtained in the latter work was probably related to a much higher moisture content of 7% wt. in the raw material, which could also explain the large water content on the obtained bio-oil. After subtracting the content in moisture, an estimated liquid yield of around 34% is obtained from the data reported by Aguiar et al. [40], being in line with the ones reported in this work. On the other hand, the superior spouted Fig. 2. Compared product distribution yields for PS, OP and SP pyrolyzed at 500 ºC under slow conditions (10ºC/min), fast conditions (>100 ºC/s) & H3PO4- treated conditions. B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057245bed reactor technology, with lower residence times for the gases and improved mass and heat transfer make possible to achieve much larger bio-oil yields [39] than the fixed bed reactor technology used in this work. To our knowledge, no pyrolysis data of saffron petals is currently available for comparison purposes. Only a few comparable works can be found, such as that of Sriram and Swaminathan about pyrolysis of Musa balbisiana flower petals. However, they only used thermogravimetric analyses and did not provide any information about the distribution of the products [41]. Regarding the results for different biomass, SP biomass gives maximum solid yield, while PS produces higher amounts of liquid and OP a higher yield of gas. This different behavior among the biomasses can be associated with their chemical structures and ash content. The aromatic and carbon-rich structure of lignin is the main contributor to the solid fraction (i.e., char) of pyrolysis processes, followed by pectin [42] and hemicellulose [43]. The results of XRF analysis of OP and SP (Table 1) show high amounts of Ca and K, which facilitate secondary reactions in favor of higher gas formation. These elements catalyze py-rolysis reactions, and they fasten the interaction between different products causing higher gas evolution [44,45]. According to previous research, potassium (K) improves the cracking of the glucosidic unit of cellulose to lower molecular weight compounds through depolymer-ization and fragmentation, which results in more water formation, CO2, methane, acetone and acetic acid, among others [46]. It is also interesting to note that the composition of the raw material has a critical impact on the product distribution, independently of the kind of pyrolysis carried out. The solid yield of cellulose-rich pistachio shells is severely decreased in fast pyrolysis, probably as the outcome of associated cellulose depolymerization reactions proceeding at faster rate than charring and repolymerization ones. As cellulose content is lower in OP and SP, the impact on the solid yield of fast pyrolysis is less important. It is also worth noting that the secondary cracking reactions from vapors of orange peel and especially of saffron petals, which are probably behind the higher gas evolution on this feedstock, are notice-ably suppressed in fast pyrolysis, increasing the corresponding liquid yield. The impregnation of biomasses with phosphoric acid has a consid-erable effect on the thermal decomposition of the biomasses, which al-ters the yields to the different pyrolysis products. Compared to slow and fast biomass pyrolysis, bio-oil formation is strongly restricted, while the formation of solid is considerably increased. Phosphoric acid catalyzes the hydrolysis of the glycosidic linkages in hemicellulose and cellulose, and it also cleaves aryl ether bonds in lignin [47] at low temperatures. As temperature increases, the organic species formed after the hydrolysis of the biopolymers can be combined with phosphorus species to form phosphate linkages. Such bonds serve to crosslink and connect the fragments of biopolymers, effectively binding volatile matter into the carbon product and, hence, delivering a net increase in carbon yield during pyrolysis, at the cost of the liquid fraction [44,47–50]. Since the main phosphoric acid interaction with lignocellulosic biomass proceeds through cellulose, the highest increase in solid yield is achieved in cellulose-rich PS. On the other hand, the main building blocks of pectins on OP are either native or methylated α-D-galactopyranosyl acid units [51], which have low reactivity with phosphoric acid, and therefore, deliver a lower increase in solid yield. 3.2. Analysis of gas fraction Fig. 3 shows the yield fraction of gas compounds in the different types of pyrolysis for PS, OP and SP. The gases consisted mainly of carbon monoxide (CO), carbon dioxide (CO2), methane (CH4), hydrogen (H2), and small amounts of light hydrocarbons (C2-C3). Occurrence of several reactions such as decarboxylation, dehydrogenation, decarbon-ylation, and hydrocarbon cracking during the pyrolysis process give rise to evolution of these gases. CO and CO2 are related to the presence of oxygen in the biomass and are generally produced because of decomposing and rearranging of carbonyl (C–O), hydroxyl (C-OH), ether (C-O-C) and carboxylic group (COO) from holocellulose and pec-tins. These are the main components of gas fraction in all these cases. CH4 is generated due to the decomposition of methoxy (-O-CH3) and methylene (-CH2-) groups, which are found on lignin and methylated pectins, while cracking and dehydrogenation of heavier hydrocarbons, which mainly occur at higher temperatures, leads to H2 production [52, 53]. When comparing the gas amounts between the biomasses, it is seen that PS produces more CO. On the other hand, CH4, C2-C3, and H2 are obtained in higher amounts in SP, with OP pyrolysis gases showing a similar composition to SP. CO evolution is mostly connected to the presence of hemicellulose and cellulose, which is larger in PS. Conversely, CO2 evolution is related to the presence of hemicellulose and pectins, explaining the higher occurrence of this gas in the pyrolysis of OP and SP. Contrarily, the formation of C2-C3 and H2 is associated with lignin decomposition and charring reactions, with lignin being found in larger quantities in PS. These trends are strongly affected by the kind of pyrolysis performed (a)(b)(c)Fig. 3. Influence of the pyrolysis process at 500 ºC in gas composition in (a) PS, (b) OP, and (c) SP under slow (10ºC/min), fast (>100 ºC/s) & H3PO4- treated conditions. B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057246(fast or acid-treated). It has been found that fast pyrolysis leads to increasing light hydrocarbons, hydrogen, and methane when compared to slow pyrolysis. The higher contribution of CH4 and H2 in fast pyrolysis has been also observed for sugarcane bagasse at 500 ◦C [54], although scarce information regarding the effect of fast pyrolysis on C2-C3 gases can be found. The results herein reported seem to point out that the increasing trend is validated no matter the composition of the feedstock, with PS reaching the highest values of hydrogen and C2-C3 light hy-drocarbons, which are probably obtained as the outcome of ring-opening reactions from the anhydrosugars monomers formed by cellulose depolymerization. Given that fast pyrolysis leads to a larger yield of the gas fraction, it is reasonable to think that highly reactive cellulose monomers, which could have faced charring and repolymeri-zation reactions during slow pyrolysis, are actually cracked at fast py-rolysis, shifting the yield from solids to gas. Differently, gas evolution from pyrolysis of H3PO4-treated biomasses shows higher contents of hydrogen and CO (syngas) compounds. In the presence of phosphoric acid, oxygen is not only removed from biomass as carbon oxides; hydrogen is also liberated as hydrogen molecule by secondary methane formation [47]. Furthermore, the decrease of CO2 also indicates that hemicellulose, as the main contributor of CO2, is cracked to a lesser extent than cellulose, probably due to the occurrence of crosslinking reactions that are promoted by phosphoric acid [48]. Interestingly, the H2/CO molar ratio of the pyrolysis gas obtained by fast pyrolysis is as high as 3.7 for pistachio shell, while in the case of H3PO4-treated biomass pyrolysis, H2/CO ratio is 3.3 for saffron petals, turning these streams as promising candidates for syngas upgrading, enabling their use as feedstock for the direct methanol synthesis or Fischer-Tröpsch process. As observed in Fig. 3, the highest LHV values obtained from the fast pyrolysis of PS and SP are due to the formation of high amounts of light hydrocarbons, like methane. However, the evolution of these gases slightly decreased in the fast pyrolysis of OP, decreasing the LHV. Finally, the improvement in LHV observed in acid-catalyzed pyrolysis mainly come from a higher syngas (CO and H2) generation. In addition, a higher LHV value is obtained for PS in this process. 3.3. Bio oil analysis The composition of the liquid fraction recovered from the condensers related to the slow and fast pyrolysis is summarized in Fig. 4. The amount of water found in the bio-oil is between 44% and 78%, with the lowest water percentage obtained in slow pyrolysis for PS. Water con-tent is higher for OP and SP independently of the heating rate used in the pyrolysis. Fast pyrolysis enhances the formation of water for all the agricultural residues, with relative increases between 9% (PS) to 28% (SP). Contrarily, the bio-oil pH value follows the order SP>OP>PS. SP bio-oil shows the lowest acidity value, 3.8, in agreement with the lower content of organic acids in the bio-oil, (see Acids family group compo-sition in Fig. 4). The acidity of the bio-oil is mostly unaffected by the heating rate for PS and SP; however, fast pyrolysis of orange peel does decrease acidity of bio-oil (pH of 3.4 in fast pyrolysis vs 2.5 in slow pyrolysis). This finding is supported by a notable decrease in acid con-tent in the composition of the organic phase of the bio-oil. The compounds found in the organic phase of the bio-oils have been scrutinized using GC-MS, and the identified compounds have been grouped based on their functional groups, in a similar way to those re-ported in other works about pyrolysis of other lignocellulosic residues [55,56]. The detailed results are summarized in Tables S1–3. Fast pyrolysis induces some differences in the composition among the families of the compounds. In general, phenol, and ketone com-pounds increase under the fast pyrolysis process for all feedstocks, while carbohydrates & derivatives, acid and furan compounds decrease. These trends fall in line with previous research for other lignocellulosic bio-masses [19,57]. For instance, higher heating rates promote the forma-tion of levoglucosan and hydroxyacetaldehyde along with a decrease in the yields of carboxylic acids [55]. Phenol and its derivatives are the principal compounds present in the bio-oil for all biomasses, Fig. 4. These compounds are mainly derived from the thermal degradation of lignin, which contains different aro-matic units within its structure [58]. It is also important to note the large increase in phenols content obtained in SP under fast pyrolysis condi-tions. Phytochemical analysis has shown that saffron flowers are rich in antioxidant compounds like flavonols (kaempferol), flavanones, antho-cyanins, crocins and crocetin [59]. Upon slow pyrolysis, flavanoids and anthocyanins, which are phenolic compounds, are probably facing charring reactions similar to those of lignin, contributing to the solid fraction. However, fast pyrolysis enhances the decomposition of the flavonoid dimers, thus increasing the phenol pool of the liquid fraction. Orange peel, which also has a high flavonoid content [60], shows a similar, but less intense, trend. The main organic components of the bio-oil obtained under both processes for PS, are phenol derivatives, Fig. 4 & Table S1. Syringol has been found as the most abundant compound for both processes, with methoxyeugenol, 5-tert-butylpyrogallol, desaspidinol, acetophenone and acetosyringone, also found in large quantities. In the case of OP and SP, fast pyrolysis delivers a huge increase in phenol derivatives, being this increase led by the rise of the syringal-dehyde family of compounds (reaching values of 14% and 12.1%, respectively). Apart from that, OP bio-oil is also rich in resorcinol, phenol and m-, o- cresols, (taking up more than 4% of peak area). As for SP, syringaldehyde compounds are followed by phenols and catechol- type compounds, like p-, o-cresol, hydroquinone, or 2,4-xylenol. These obvious changes in the aromatic product distribution must be related to the different nature of lignin between samples. Lignin in PS probably has a similar composition to those observed in hardwoods, based on syringol and guaiacol units; lignin in OP is probably mainly constituted by guaiacol and p-hydroxyphenyl units, meanwhile SP lignin should be composed mostly by the latter ones. The use of high heating rates seems to promote lignin depolymerization over cracking and charring reactions in all biomasses, explaining the huge increase in phenols derived from the monolignols. The presence of inorganic matter can also promote the charring reactions of lignin in SP and OP. When the heating rate increases, the slow crosslinking and condensation reactions that are catalyzed by the inorganic elements cannot take place, and monolignols and their derived phenols are released into the gas phase, explaining the large increase of these compounds in the fast pyrolysis bio-oils of SP and OP. The phenol and its derivatives are valuable and useful as a resin, to produce antioxidants, dye, and pharmaceuticals [61]. Acid compounds constitute a family unavoidably present in bio-oil. Fig. 4. Analysis of the pyrolysis liquid phase for three biomasses under slow and fast pyrolysis. B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057247Acids in the bio-oil fraction are attributed mainly to the decomposition of the acetyl groups present in hemicellulose [62] and pectin [63]. Acetic acid is the main compound in this group, whose concentration decreases significantly under the fast pyrolysis of all the biomasses. The amount of acids present in bio-oil of each biomass is consistent with the pH results, Fig. 4. Unfortunately, acidity is one of the main problems of bio-oil and its recovery by separation from bio-oil would allow their valorization as a by-product and would ease the use and upgrading of the acid-free bio-oil. Another important family found in bio-oil is furan. Furanic com-pounds, especially furfural and 5-hydroxymethylfurfural are recognized as dehydration products of sugars [53]. The relative amounts found in this study are quite dependent on the feedstock and pyrolysis process. In this sense, SP pyrolysis produces the highest furan concentrations, whereas fast pyrolysis decreases their concentration for all residues. SP bio-oils are rich in 2(5 H)-furanone, 2,3-dihydrobenzofuran, furfuryl alcohol, 5-hydroxymethylfurfural, and dihydro-4-hydroxy-2 (3 H)-furanone. Differently, PS bio-oil is rich in furfural, while furfuryl alcohol, furfural, 5-hydroxymethylfurfural and 5-methyl-2-furaldehyde are detected in the liquid phase obtained from OP. Tables S1–3 show a decline in branched furans in bio-oils derived from fast pyrolysis, probably as a consequence of the promotion of secondary cracking re-actions. In slow pyrolysis, hemicellulose decomposition and furan evo-lution take place at mild temperatures (250–350 ºC), avoiding the occurrence of secondary cracking reactions that probably take place under fast pyrolysis conditions at 500 ºC. It has been found that fast pyrolysis promotes the formation of ke-tone compounds, with OP being the most suitable agricultural residue for the obtention of ketones, Fig. 4. The increase observed in fast py-rolysis seems to be connected to a larger formation of cyclopentanone compounds, a tendency that is confirmed for all the feedstocks no matter their biopolymeric composition, Tables S1–3. The ketone compounds are usually derived from hemicellulose and cellulose degradation, and it has been reported that they are formed during pyrolysis through the condensation of carbohydrates and decomposition of oxygenate com-pounds and furans due to secondary ring scission and deacetylation reactions of sugars and furans [64]. Compounds of this family can be also grouped according to their linear or cyclic structure. The formation of cyclopentanone, the most representative compound in the second group, has been attributed to hemicellulose decomposition [65], with the cyclic ketones being likely derived from the xylan present in stem by cleavage of the o-glucosidic bonds and subsequent removal of the hy-droxyl groups of the xylose rings [66]. Another feature observed in the obtained bio-oil from fast pyrolysis is the conversion of aldehyde com-pounds into cyclic ketones, as can be seen in Tables S1–3. Some saccharides are detected in PS and SP bio-oils. The maximum amount of this family is related to D-Allose, 4%, which appeared in PS bio-oil obtained under fast pyrolysis, Table S1. Beyond D-Allose, maltols are the next main component of bio-oils from OP and SP. These com-pounds are suggested to be obtained via dehydration and rearrangement of levoglucosan and 1,4:3, 6-dianhydro-α-D-glucopyranose, which is known to happen at low temperatures (200 ºC) during the pyrolysis of sucrose or starch [67]. Their low contribution to the bio-oil fraction can be related to their low thermal stability, which favors their decompo-sition into different oxygenate compounds, such as furans, ketones and aldehydes. Accordingly, these compounds are likely degraded under fast pyrolysis conditions, explaining the decline observed in Fig. 4. The amounts of N-containing compounds are in the order SP>OP>PS, consistent with the nitrogen content reported in the elemental analyses of the feedstocks, Table 1. According to previous research, the formation of these compounds is related to the interaction between proteins and carbohydrates. Amino acids are usually more reactive than other compounds, being prone to react with active cellu-lose and derived compounds such as hydroxypropanone. For instance, pyrolysis of glutamic acid in lignocellulosic biomass leads to pyrrolidi-none compounds [68], which are found in orange peel. Pyrolysis of proteoglycans at temperatures over 500 ºC can also generate pyridines, pyrroles, and similar N-aromatic compounds [65]. The generation of compounds such as methyl L-pyroglutamate, i.e., the main nitrogen-containing specie found in SP, Table S3, has been reported for direct pyrolysis of amino acids, meaning that protein concentration should be rather high in saffron petals for enabling their formation, as reported elsewhere [69]. Finally, the production of N-compounds has been also reported as feasible in raw materials with low N-content, such as cellulose and cellobiose [70], obtaining oxazolidine and other com-pounds resembling those observed in PS bio-oil, Table S1, but in very low amounts. Therefore, the formation of N-derived compounds is largely dependent on the presence of proteins, proteoglycans, amino acids and other N-containing species in the feedstock [39,71]. It should be pointed out that the bio-oil obtained under fast pyrolysis has much lower concentration of nitrogen compounds. It seems like amino acids decomposed during fast pyrolysis could be incorporated into the char forming pyrroles, pyridines, and quaternary nitrogen [72]. In fact, elemental analysis of SP char obtained by fast pyrolysis confirms the presence of nitrogen, reaching 2.8% wt. The presence of such high ni-trogen concentration can be of interest for certain electrochemical ap-plications [73]. Generally, after upgrading the bio-oil can be used in different application such as biofuel or feedstock for oxygenates, hy-drocarbons (olefins/aromatics) or hydrogen, or application in the pro-duction of carbon materials, asphalts, pesticides, fertilizers, perfumes, polyurethane foam or plastic [74]. Table 2 summarizes the detected compounds in the liquid phase obtained from H3PO4-treated biomass pyrolysis together with their water content and pH values. When the biomasses are treated with H3PO4, the obtained liquid phase is completely different from those of slow and fast pyrolysis, starting by the clearer, aqueous-like appearance. As previously discussed, condensation reactions catalyzed by H3PO4 proceed particularly through dehydration reactions. It is also important to remark that phosphoric acid contains 15% wt. of water. Thus, a huge increase in water content is found in these bio-oils, Table 2. Moreover, the pH value is slightly increased (see the differences in pH in Fig. 4), since, probably, part of phosphorus leaves the reactor as phosphoric acid ester compounds (as pointed out by the presence of phosphoric acid, diethyl pentyl ester, Table 2), which are hydrolyzed in the bio-oil, increasing acidity. In addition, this process seems to demote the release of organic volatiles, with the liquids consisting of only a few Table 2 Composition analysis of liquid-phase obtained from H3PO4 acid treatment. Compounds Biomass PS OP SP Amount (area %) Acetic acid, butyl ester 85.9 48.6 35.6 Acetic acid 4.6 10.5 17.8 Toluene 4.1 20.1 16.4 Phenol 4.3 10.2 14.4 Phenol, 4-methyl- 0.0 4.9 6.8 Phenol, 3-methyl- 0.0 2.1 2.8 3-(Methoxymethoxy)butanoic acid 0.0 0.0 2.7 Phosphoric acid, diethyl pentyl ester 1.1 3.6 2.0 Furan, 2-ethyl-5-methyl- 0.0 0.0 1.4 pH 1.3 2.4 2.6 Water (%wt.) 88 83 88 B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057248particular products. It is important to note that the concentration of these compounds is low, since the liquid fraction obtained in acid treated pyrolysis has less than 15% in organic compounds (see water composition in Table 2). As can be seen in Table 2, acetic acid and acetic acid derived esters are the main compounds found in the three bio-masses. Next to them, the aromatic family of compounds, with toluene, phenol, and phenol derivatives as the major chemicals, are detected in large amounts in SP and OP bio-oils. Interestingly, ketone, furans, car-bohydrates and N-containing compounds are not found in relevant amounts in the bio-oil, pointing out that phosphoric acid is promoting the incorporation of cellulose, hemicellulose and amino acids -derived compounds in the solid fraction of the pyrolysis products. Moreover, these results are also showing that the selectivity of the products in the bio-oils obtained from acid treated biomass is somehow improved. In this sense, Lobos et al. pyrolyzed a Kraft pulp waste with H3PO4. Their selectivity to bio-oils was particular and they claimed that only levo-glucosenone (LGO) was achieved [75]. The higher amount of H3PO4 and higher pyrolysis temperature used in this work seem to shift the selec-tivity towards different products. 3.4. Analysis of solid fraction The ultimate and proximate analyses and the higher heating values (HHVs) of the solid fraction for each biomass obtained from different types of pyrolysis are compiled in Table 3. Fixed carbon increases under fast pyrolysis, while moisture, volatile material, and ash contents decrease compared to slow pyrolysis. Acid treatment gives rise to solids with higher moisture and fixed carbon contents and lower volatile and ash amounts. As for proximate analysis, the solids show similar carbon, nitrogen, and sulfur composition independently of the heating rate used, but fast pyrolysis provides a slightly higher amount of oxygen [18]. It is also interesting to note that hydrogen content declines with phosphoric acid treatment in OPC and SPC solids, confirming the higher H2 releases observed in Fig. 3. The presence of high amounts of nitrogen, oxygen and ash in the solid residues from OP and SP makes feasible their possible use as raw material for the preparation of catalysts and soil improvers [76–78]. For instance, Yao et al. studied the performance of a Ni/ char catalyst for biomass gasification, achieving a high H2 yield of 64 vol%, 92.08 mg g− 1 biomass at the optimum operation conditions [77]. HHV of the solids has been determined from their proximate analysis using Eq. (2). The obtained values are in the range of 21–33.5 MJ/kg. According to Table 3, the maximum HHVs are obtained under the py-rolysis with H3PO4-treated biomasses, which is due to the removal of large quantities of ashes during the washing step. Differently, fast py-rolysis has a low impact on HHV. Comparing the different feedstocks, the best results in terms of HHV are obtained for solids from PS, due to the higher amount of fixed carbon and lower amount of ashes compared to their counterparts. The energy content of OP solids shows medium values, like those reported for charred wood, while PS solids are in the upper range of HHVs, being close to those reported for biochar obtained from straw [79]. Finally, SP shows poorer HHV values owing to the high amount of inorganic matter (i.e., ash content) in their composition. Fig. 5 shows the N2 adsorption–desorption isotherms at − 196 ◦C for the three solid fractions obtained from acid treated biomass pyrolysis. The N2 uptake of the solid residues (chars) from slow and fast pyrolysis are negligible due to the poor and narrow microporosity development, which is not accessible for N2 adsorption at − 196 ºC, and therefore they are not included. In general, all the N2 adsorption isotherms reported in Fig. 5 shows notable N2 uptake at very low relative pressures followed by monoto-nous increase in the quantity of N2 adsorbed as relative pressure in-creases, which are regarded as type I + type IV isotherms according to IUPAC classification [37], corresponding to a well-developed micropo-rous structure with a significant contribution of mesoporosity. The isotherm from the SPC sample presents the lowest N2 uptake at low relative pressures and the lowest slope of the isotherm at medium and high pressures. The appearance of small H4 type hysteresis loop (no limited adsorption at relative pressures close to 1) indicates the presence of narrow slit-shaped pores. The OPC and, especially, PSC show a significantly higher contribution of mesoporosity, as revealed by the higher N2 volume adsorbed at medium-high relative pressures and the larger hysteresis loops. In the case of PSC, H1 type hysteresis with certain contribution of H4 is observed, indicative of narrow mesopores, while the H2 type hysteresis loop of OPC (i.e. steeper slope of the desorption with respect to the adsorption isotherms) suggests the pres-ence of a broad mesopore size distribution of varied shapes [37]. The textural parameters of the porous structure of the different Table 3 Ultimate and proximate analysis and HHV values of chars for all the pyrolysis processes. Sample Proximate analysis (wt%) Ultimate analysis (dafa, wt%) HHV (MJ/Kg, dry basis) M VM FC ash C H N S Ob PS (slow) 4.0 23.3 70.4 2.3 88.5 3.1 0.1 0.1 8.2 30.1 OP (slow) 8.5 28.0 50.7 12.8 78.2 3.5 2.2 0.0 16.1 24.9 SP (slow) 6.5 24.7 44.2 24.6 82.1 3.3 3.2 0.8 10.6 21.3 PS (fast) 2.8 15.9 79.5 1.8 87.3 3.4 0.1 0.0 9.2 31.8 OP (fast) 5.6 24.9 54.0 15.5 77.2 3.8 2.3 0.1 16.6 24.8 SP (fast) 4.4 23.1 52.0 20.5 79.9 3.2 2.8 0.1 14.0 23.4 PSC 12.2 5.9 80.1 1.8 94.8 2.5 0.2 0.0 2.5 33.5 OPC 10.8 8.1 70.5 10.6 84.1 1.9 0.4 0.0 13.6 30.5 SPC 6.5 5.2 77.7 10.6 86.2 1.7 2.8 0.2 9.1 30.4 M=moisture, VM= volatile material, FC= fixed carbon. a dry ash free. b calculated by difference. Fig. 5. N2 adsorption–desorption isotherms at − 196◦C of solid residues ob-tained from the acid treated biomass pyrolysis. B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 1057249solids, obtained from the N2 and CO2 isotherms, are summarized in Table 4. In the case of slow and fast pyrolysis, all the solids have VCO2>VN2 (Table 4), with the largest CO2 micropore volume being attained in PS char obtained at slow pyrolysis. This difference between isotherms can be explained by diffusional limitations of N2 during adsorption at − 196 ◦C in narrow micropores (those with sizes under 0.5–0.7 nm). Even though the CO2 molecule presents a similar kinetic diameter to N2, the temperature for the determination of the CO2 adsorption isotherm, 0 ◦C, boosts the diffusion rate in narrower micropores, enabling it to reach the adsorption equilibrium in shorter times. Thus, all the slow and fast pyrolysis solids have narrow microporosity. Fast pyrolysis seems to hamper the porosity development, probably due to the suppression of decomposition reactions of functional groups and rearrangement of the carbonaceous matrix of the solid when pyrolysis takes place at fast heating rate. In the case of acid treated biomass, different results are attained, and the porosity is vigorously developed. The micropore volume values obtained from N2 and CO2 adsorption isotherms, with VDRN2 > VDRCO2, indicate the presence of wide microporosity in these solids (sizes larger than 0.7 nm) [80]. In terms of surface area (ABET) and total pore volume (V0.99), the porosity development of the solid fractions follows the sequence: PSC > OPC > SPC, probably, due to the presence of higher cellulose in PS [81–83]. Higher surface areas in activated carbons pro-duced from cellulose than those of Kraft lignin are already reported in the work of Bedia et al. [82], which showed that the raw materials can have an enormous impact on porosity development. The ABET value obtained for PSC solids with 1640 m2/g, is comparable to those acti-vated carbons obtained by chemical activation of hemp resides or pine dust with H3PO4, and much higher than those attained by H3PO4 treatment of lignin or cellulose alone [44,84]. The total pore volume of PSC is also the highest one, equal to 1.41 cm3/g. In addition, Vmeso/V0.99 ratios of 0.701, 0.631, and 0.542 are obtained for OPC, PSC, and SPC solids, pointing out that the use of orange peel as raw material delivers wider mean porosity. This trend is confirmed by the average pore vol-ume as estimated by 4⋅VT/ABET, being 4.5 nm for OPC vs 3.4 and 3.2 nm for PSC and SPC, respectively. However, when the pore size distribution is analyzed in detail, more differences arise. The NLDFT pore size dis-tribution of the solid fractions from acid treated biomass pyrolysis is plotted in Fig. 6. Two common features are found in the PSDs of all the materials, namely i) narrow distribution of micropores (2 <nm) showing a maximum at 0.6 nm ii) broad distribution of mesopores running up to 20 nm. The position and extension of the PSD in the mesopore region are different. In the case of SPC solid, the mesopore size distribution ends at 10 nm and it does not show a distinctive maximum, pointing out the presence of narrow mesoporosity. PSC distribution of mesopores is centered at 2.6 nm, while OPC shows the broadest meso-pore size distribution, with a maximum at 4.7 nm. The generation of porosity can be related to the chemical reactions between H3PO4 and carbon precursors. The presence of C–O–PO3 sur-face groups on some porous carbons prepared by chemical activation of biomass residues with H3PO4 at different temperatures was confirmed by XPS analyses [42,79,80]. These phosphates and/or polyphosphates groups seems to be responsible for the dilation process during activation [47], and they have a significant impact not only on the porous texture but on the surface chemistry of activated carbons, providing surface acidity [85], oxidation [86], and electrooxidation [87] resistance. Thus, the combination of a large surface area with the presence of mesopores and phosphorus groups make OPC and PSC solid fractions great candi-dates for their use as catalytic support and adsorbents in pollutant remediation [82,85,88–90]. In this sense, Valero-Romero et al. prepared phosphorus-containing mesoporous carbon from H3PO4 activation with olive stone that was successfully used as catalysts for methanol dehy-dration [88]. Additionally, H3PO4 modified biochar has been applied for removing toxic Cr (VI) by Nawaz et al. The results showed that chemi-cally modified biochar developed a moderate amount of surface area (246 m2g− 1) along with microporosity, showing Cr (VI) removal of 99.97% under optimized conditions [89]. In another study about the remediation of emergent pollutants, such as bisphenol A and carba-mazepine, the use of a H3PO4-treated biochar showed a clear increase in the adsorption capacities [90]. 4. Conclusions The results obtained from the fixed-bed pyrolysis of pistachio shells, bitter orange peels and saffron petals under slow heating rate, fast heating rate and H3PO4 treatment conditions are herein reported. Gas fraction of the fast pyrolysis is rich in C2-C3, CH4 and H2, increasing HHV. The H3PO4 treatment of biomasses leads to the evolu-tion of higher syngas (CO and H2), which also increases the HHV compared to slow pyrolysis. The composition of the gas fraction is affected by the presence of hemicellulose, pectins and inorganic matter in the feedstocks, which render a higher formation of CO2, decreasing the HHV. GC-MS results shows that the bio-oil obtained under fast and slow pyrolysis are richer in phenolic and acidic functional groups. However, fast pyrolysis promotes the formation of water, phenols, and ketones, while hindering the generation of N-compounds, carbohydrates and derivatives, acid, and furan compounds in the bio-oil. The bio-oil ob-tained from acid treated-biomass pyrolysis shows water as the main compound, with acids and aromatic compounds as the main components of the organic fraction. Likewise, the HHV values of the solid residues obtained from fast pyrolysis are higher than those of slow pyrolysis. Acid catalyzed pyrol-ysis further enhances the HHVs. Additionally, among the biomasses, the PS solid residues show the highest HHVs, which is mostly related to the presence of lower inorganic matter. Moreover, SP and OP solid residues Table 4 Characteristic parameters of the porous texture for solid residues under different pyrolysis conditions. Sample ABET m2 g− 1 VDRN2 cm3 g− 1 Vmeso cm3 g− 1 V0.99 cm3 g− 1 VDRCO2 cm3 g− 1 PSC 1640 0.52 0.89 1.42 0.26 OPC 870 0.29 0.68 0.97 0.16 SPC 300 0.11 0.13 0.25 0.07 PS (slow) 1.1 < 0.01 < 0.01 < 0.01 0.19 OP (slow) 1.0 < 0.01 < 0.01 0.01 0.11 SP (slow) 0.6 < 0.01 < 0.01 < 0.01 0.11 PS (fast) 1.0 < 0.01 < 0.01 < 0.01 0.08 OP (fast) 0.9 < 0.01 < 0.01 < 0.01 0.08 SP (fast) 0.9 < 0.01 < 0.01 < 0.01 0.07 Fig. 6. Pore distribution of the solid residues obtained from acid treated biomass pyrolysis. Inset: micropore region. B. Hosseinzaei et al. Journal of Analytical and Applied Pyrolysis 168 (2022) 10572410present certain amounts of nitrogen in their composition, which could be valuable for the preparation of nitrogen-doped activated carbons. Solid residues obtained from acid treated biomass pyrolysis develop the porosity, being comparable to the values of apparent surface area of commercial activated carbons. The maximum porosity was achieved for PSC solid residue, with ABET of 1640 m2/g together with a total pore volume of 1.42 cm3/g. CRediT authorship contribution statement Behnam Hosseinzaei: Methodology, Investigation, Visualization, Writing – original draft. Mohammad Jafar Hadianfard: Conceptuali-zation, Validation, Supervision. Ramiro Ruiz-Rosas: Investigation, Visualization, Formal analysis, Data curation, Writing – review & edit-ing. Juana M. Rosas: Data curation, Writing – review & editing. José Rodríguez-Mirasol: Conceptualization, Supervision, Project adminis-tration, Funding acquisition. Tomás Cordero: Conceptualization, Su-pervision, Project administration, Funding acquisition. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data availability Data will be made available on request. Acknowledgments RRR, JMR, JRM and TC thank MICINN (RTI2018-097555-B-100) and Junta de Andalucía (P18-RT-4592) for financial support. Funding for open access charge: Universidad de Málaga/CBUA. Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jaap.2022.105724. References [1] H.L. Chum, R.P. Overend, Biomass and renewable fuels, Fuel Process. Technol. 71 (1–3) (2001) 187–195, https://doi.org/10.1016/S0378-3820(01)00146-1. [2] Y. Liao, S. Koelewijn, G. van den Bossche, J. Gvan Aelst, S. van den Bosch, T. Renders, K. Navare, T. Nicolaï, K. van Aelst, M. Maesen, H. Matsushima, J. Thevelein, K. van Acker, B. Lagrain, D. Verboekend, B. Sels, A sustainable wood biorefinery for low-carbon footprint chemicals production, Science 367 (6484) (2020) 1385–1390, https://doi.org/10.1126/science.aau1567. [3] A. Mlonka-Mędrala, P. Evangelopoulos, M. Sieradzka, M. Zajemska, A. Magdziarz, Pyrolysis of agricultural waste biomass towards production of gas fuel and high- quality char: experimental and numerical investigations, Fuel 296 (2021), https:// doi.org/10.1016/j.fuel.2021.120611. [4] A. Taghizadeh-Alisaraei, H.A. Assar, B. Ghobadian, A. Motevali, Potential of biofuel production from pistachio waste in Iran, Renew. Sustain. Energy Rev. 72 (2017) 510–522, https://doi.org/10.1016/J.RSER.2017.01.111. [5] IndexBox: Volume of World Pistachio Market in 2019 Grew By 4%, EastFruit. 〈htt ps://east-fruit.com/en/horticulture-market/market-reviews/indexbox-volume-o f-world-pistachio-market-in-2019-grew-by-4/〉, (Accessed 24 June 2022). [6] L. Navarro, The Spanish citrus industry, Acta Hortic. (2015) 41–48, https://doi. org/10.17660/ActaHortic.2015.1065.1. [7] Food and Agriculture Organization of the United Nations. 〈https://www.fao. org/faostat/en/#home〉, (Accessed 21 June 2022). [8] Saffron Production in 2019, Alipour Inc. (n.d.). 〈https://alipourinc.com/saffron- production-in-2019/〉, (Accessed 1 February 2022). [9] F.X. Collard, J. Blin, A review on pyrolysis of biomass constituents: Mechanisms and composition of the products obtained from the conversion of cellulose, hemicelluloses and lignin, Renew. Sustain. Energy Rev. 38 (2014) 594–608, https://doi.org/10.1016/J.RSER.2014.06.013. [10] T.Y.A. Fahmy, Y. Fahmy, F. Mobarak, M. El-Sakhawy, R.E. Abou-Zeid, Biomass pyrolysis: past, present, and future, Environ. Dev. Sustain. 22 (1) (2020) 17–32, https://doi.org/10.1007/s10668-018-0200-5. [11] S. Yaman, Pyrolysis of biomass to produce fuels and chemical feedstocks, Energy Convers. Manag. 45 (5) (2004) 651–671, https://doi.org/10.1016/S0196-8904 (03)00177-8. [12] S. Fiore, F. Berruti, C. Briens, Investigation of innovative and conventional pyrolysis of ligneous and herbaceous biomasses for biochar production, Biomass Bioenergy 119 (2018) 381–391, https://doi.org/10.1016/J. BIOMBIOE.2018.10.010. [13] E. Fernandez, L. Santamaria, M. Amutio, M. Artetxe, A. Arregi, G. Lopez, J. Bilbao, M. Olazar, Role of temperature in the biomass steam pyrolysis in a conical spouted bed reactor, Energy 238 (2022), 122053, https://doi.org/10.1016/j. energy.2021.122053. [14] J.L. Toro-Trochez, D.A. De Haro Del Río, L. Sandoval-Rangel, D. Bustos-Martínez, F.J. García-Mateos, R. Ruiz-Rosas, J. Rodríguez-Mirasol, T. Cordero, E.S. Carrilo- Pedraza, Catalytic fast pyrolysis of soybean hulls: focus on the products, J. Anal. Appl. Pyrolysis (2022) 163, https://doi.org/10.1016/j.jaap.2022.105492. [15] D. Mohan, C.U. Pittman, P.H. Steele, Pyrolysis of wood/biomass for bio-oil: a critical review, Energy Fuels 21 (2006) 848–889, https://doi.org/10.1021/ ef0502397. [16] A. Pattiya, Bio-oil production via fast pyrolysis of biomass residues from cassava plants in a fluidised-bed reactor, Bioresour. Technol. 102 (2) (2011) 1959–1967, https://doi.org/10.1016/j.biortech.2010.08.117. [17] R.M. Taib, N. Abdullah, N.S.M. Aziz, Bio-oil derived from banana pseudo-stem via fast pyrolysis process, Biomass Bioenergy 148 (2021), 106034, https://doi.org/ 10.1016/J.BIOMBIOE.2021.106034. [18] G. Duman, C. Okutucu, S. Ucar, R. Stahl, J. Yanik, The slow and fast pyrolysis of cherry seed, Bioresour. Technol. 102 (2) (2011) 1869–1878, https://doi.org/ 10.1016/J.BIORTECH.2010.07.051. [19] Z. Yang, A. Kumar, R.L. Huhnke, M. Buser, S. Capareda, Pyrolysis of eastern redcedar: distribution and characteristics of fast and slow pyrolysis products, Fuel 166 (2016) 157–165, https://doi.org/10.1016/J.FUEL.2015.10.101. [20] R. Verma, S. Verma, S. Verma, J. Wang, J. Liu, B. Jing, K. Rakesh, Value-addition of wheat straw through acid treatment and pyrolysis of acid treated residues, J. Clean. Prod. 282 (2021), 124488, https://doi.org/10.1016/J.JCLEPRO.2020.124488. [21] H. Wang, R. Srinivasan, F. Yu, P. Steele, Q. Li, B. Mitchell, A. Samala, Effect of acid, steam explosion, and size reduction pretreatments on bio-oil production from sweetgum, switchgrass, and corn stover, Appl. Biochem. Biotechnol. 167 (2012), https://doi.org/10.1007/s12010-012-9678-8. [22] Z.L.L. Cueva, G.J. Griffin, L.P. Ward, S. Madapusi, K.V. Shah, R. "}}}],"truncated":false,"partial":false},"paginationData":{"pageIndex":0,"numItemsPerPage":100,"numTotalItems":100,"offset":0,"length":100}},"jwt":"eyJhbGciOiJFZERTQSJ9.eyJyZWFkIjp0cnVlLCJwZXJtaXNzaW9ucyI6eyJyZXBvLmNvbnRlbnQucmVhZCI6dHJ1ZX0sImlhdCI6MTc3OTA1MjI4Niwic3ViIjoiL2RhdGFzZXRzL3RvdXJpc3Q4MDAvU0NJLUlERUEiLCJleHAiOjE3NzkwNTU4ODYsImlzcyI6Imh0dHBzOi8vaHVnZ2luZ2ZhY2UuY28ifQ.wVpSnYZMBmpkYl0fHe4kZNYUCiLCaEo6K0xhHlMXi1kpcNojLhQx4XfcxyD715PxAOZp_iy6aRxNJUgg2WU2Bg","displayUrls":true,"splitSizeSummaries":[{"config":"default","split":"train","numRows":100,"numBytesParquet":3218651}]},"discussionsStats":{"closed":0,"open":1,"total":1},"fullWidth":true,"hasGatedAccess":true,"hasFullAccess":true,"isEmbedded":false,"savedQueries":{"community":[],"user":[]}}">
Researcher Name stringlengths 5 28 | ORCID stringlengths 1 23 | Researcher Query Keyword stringlengths 22 96 ⌀ | Research Full Paper stringlengths 2 65.5k | Similar Full Paper stringlengths 2 65.5k |
|---|---|---|---|---|
Daniel Halpern | - | Aligning AI with Human Values via RLHF and Social Choice Theory | {'Metric Distortion with Elicited Pairwise Comparisons': 'Title: Metric Distortion with Elicited Pairwise Comparisons\\nAbstract\\nIn this work we study the metric distortion problem in voting theory under a limited\\namount of ordinal information. Our primary contribution is threefold. First, we consider\\nmechanisms ... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Zhen Xiang | -,0009-0002-9077-2399,- | BadChain: Backdoor Attacks on Chain-of-Thought Prompting in LLMs | {'Combining Structural Knowledge with Sparsity in Machine Learning and Signal Processing': 'Title: Combining Structural Knowledge with Sparsity in Machine Learning and Signal Processing\\nAbstract\\nBayesian networks represent relations between variables using a directed acyclic graph (DAG). Learning\\nthe DAG is an NP... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Zhixiang Shen | - | Multiplex Graph Fusion | {'Generalized Channel Coding and Decoding With Natural Redundancy in Protocols': 'Title: Generalized Channel Coding and Decoding With Natural Redundancy in Protocols\\nAbstract—In many wireless systems, the signal-to-interference-\\nand-noise ratio that is applicable to a certain transmission,\\nreferred to as channel ... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Matthew Thomas Jackson | - | Learned Optimization for Reinforcement Learning | {'Can Learned Optimization Make Reinforcement Learning Less Difficult?': 'Title: Can Learned Optimization Make Reinforcement Learning Less Difficult?\\nAbstract\\nDeep reinforcement learning has achieved many\\nrecent successes, but our understanding of its\\nstrengths and limitations is hampered by the lack\\nof rich ... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Yiyang Zhao | - | Carbon-Efficient Neural Architecture Search (CE-NAS) | {'Carbon-Efficient Neural Architecture Search (CE-NAS)': 'Title: Carbon-Efficient Neural Architecture Search (CE-NAS)\\nCarbon-Efficient Neural Architecture SearchYiyang ZhaoWorcester Polytechnic Instituteyzhao10@wpi.eduTian GuoWorcester Polytechnic Institutetian@wpi.eduABSTRACTThis work presents a novel approach to ne... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Denny Zhou | - | Inducing Chain-of-Thought Reasoning via Decoding Adjustments | {'Chain of Thought Empowers Transformers to Solve Inherently Serial Problems': 'Title: Chain of Thought Empowers Transformers to Solve Inherently Serial Problems\\nAbstract Identifying pathways for small to medium sized enterprises (SMEs), as potential entrants to the international market place, encourages evaluation ... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Siyuan Huang | - | Cluster-wise Graph Transformer | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Noah A. Smith | 0000-0002-2310-6380 | Multi-Objective Language Model Alignment | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Simon S. Du | 0000-0003-0056-8299 | Multi-Objective Language Model Alignment | {'Decoding-Time Language Model Alignment with Multiple Objectives': 'Title: Decoding-Time Language Model Alignment with Multiple Objectives\\nabstract objectives such asharmlessness and helpfulness (proposed in thepost-LLM era) show that we can DeAL withfine-grained trade-offs, improve adherenceto alignment objectives,... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Ke Sun | - | Enhanced Deepfake Detection with Diffusion Models | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Hong Liu | 0000-0001-5318-6388 | Enhanced Deepfake Detection with Diffusion Models | {'Enhanced Deepfake Detection with Diffusion Models': 'Title: Enhanced Deepfake Detection with Diffusion Models\\nAbstract Mean-ing Representation of the caption and then for-wards the query-image pairs into a pre-trainedlarge vision-language model select the “evi-dences" that are helpful for us to detect mis-informati... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Yiming Yang | - | Easy-to-Hard Generalization in AI Alignment | {'Fast Closed Tree Clustering Parallel Algorithm for Dynamic Cloud Platform': 'Title: Fast Closed Tree Clustering Parallel Algorithm for Dynamic Cloud Platform\\nAbstract\\nThere has been significant recent interest in parallel graph processing due to the need to\\nquickly analyze the large graphs available today. Many... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Chuang Gan | - | Easy-to-Hard Generalization in AI Alignment | {'Improving Reinforcement Learning from Human Feedback with Efficient Reward Model Ensemble': 'Title: Improving Reinforcement Learning from Human Feedback with Efficient Reward Model Ensemble\\nImproving the generalizability and robustness of large-scale traffic signal controlTianyu ShiDepartment of Civil Engineering, ... | {'Learning to Reason via Self-Iterative Process Feedback for Small Language Models': "Title: Learning to Reason via Self-Iterative Process Feedback for Small Language Models\\nAbstract \\t vi \\nStatement of authorship \\t vii \\nAuthority of Access Statement \\t vii \\nDedication and Acknowledgement viii \\nChapter 1 ... |
Roman Klinger | 0000-0002-2014-6619 | Cross-Lingual Emotion Classification | {'Natural Language Inference Prompts for Zero-shot Emotion Classification in Text across Corpora': 'Title: Natural Language Inference Prompts for Zero-shot Emotion Classification in Text across Corpora\\nabstract 4 Experimentsemotion-related names leads to a better model.Hence, we set r(e) to return a set of emotionsyn... | {'Document-Level Sentiment Analysis of Urdu Text Using Deep Learning Techniques': 'Title: Document-Level Sentiment Analysis of Urdu Text Using Deep Learning Techniques\\nAbstract\\nSisi LIU\\nDocument-level Sentiment Analysis of Email Data\\nWith the increasing prevalence of electronic devices and advances in network\\... |
Jungmo Kim | - | Offline Reinforcement Learning | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Dennis G. Wilson | 0000-0003-2414-0051 | Skill Diversity in Reinforcement Learning | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... | {'Comprehensive Study on Reinforcement Learning and Deep Reinforcement Learning Schemes': 'Title: Comprehensive Study on Reinforcement Learning and Deep Reinforcement Learning Schemes\\nAbstract.Deep learning basedmodels have surpassed classical machine learning based approaches in various text classification\\ntasks, ... |
Alyssa Hwang | - | Multi-hop Question Answering | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... | {'Unraveling the Capabilities of Language Models in News Summarization': 'Title: Unraveling the Capabilities of Language Models in News Summarization\\nThread of Thought Unraveling Chaotic ContextsYucheng Zhou1∗, Xiubo Geng2, Tao Shen3, Chongyang Tao2,Guodong Long3, Jian-Guang Lou2†, Jianbing Shen1†1 SKL-IOTSC, CIS, Un... |
Jennifer D'Souza | 0000-0002-6616-9509 | Cognitive Knowledge Graphs for Scholarly Article Categorization | {'Scholarly Question Answering using Large Language Models in the NFDI4DataScience Gateway': 'Title: Scholarly Question Answering using Large Language Models in the NFDI4DataScience Gateway\\nabstract, author, and datePublishedinherited by its subclasses. Each class within schema.org contributes to a struc-tured repres... | {'Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation': 'Title: Expanding the Scope: Inductive Knowledge Graph Reasoning with Multi-Starting Progressive Propagation\\nAbstract. Knowledge graphs (KGs) are widely acknowledged as incom-\\nplete, and new entities are constan... |
Dawei Cheng | 0000-0002-5877-7387 | Graph Training Debugging with Dynamic Soft-Pruning | {'Dynamic Graph-based Deep Reinforcement Learning with Long and Short-term Relation Modeling for Portfolio Optimization': 'Title: Dynamic Graph-based Deep Reinforcement Learning with Long and Short-term Relation Modeling for Portfolio Optimization\\nAbstract\\nGraphs arise naturally in many real-world applications incl... | {'Intelligent system for anomaly detection and decision-making support based on Semantic Web technologies in manufacturing processes in Aerospace Industry (short paper)': 'Title: Intelligent system for anomaly detection and decision-making support based on Semantic Web technologies in manufacturing processes in Aerospa... |
Binghong Wu | - | Multimodal Generative Modeling | {'Harmonizing Visual Text Comprehension and Generation': 'Title: Harmonizing Visual Text Comprehension and Generation\\nABSTRACT Harmonizing Music and Language Acquisition: Learning Language with the Arts by Samantha Jay Ontiveros: Master of Second Language Teaching Utah State University, 2020 Major Professor: Dr. ... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Emmanuel Anaya Gonzalez | 0009-0002-9013-2228 | LLM-Guided Program Synthesis | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Roopal Garg | 0009-0001-3343-9226 | Human-in-the-loop Image Description | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Extract Free Dense Misalignment from CLIP': 'Title: Extract Free Dense Misalignment from CLIP\\nAbstract .............................................................................................. 1', 'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelih... |
Burcu Karagol Ayan | - | Human-in-the-loop Image Description | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Zhengyan Shi | - | Instruction-Enhanced Language Model Training | {'Instruction Tuning With Loss Over Instructions': 'Title: Instruction Tuning With Loss Over Instructions\\nContrastive Instruction TuningTianyi Lorena Yan♢, Fei Wang♢, James Y. Huang♢, Wenxuan Zhou♢, Fan Yin♠Aram Galstyan♢, Wenpeng Yin♡, Muhao Chen♣♢University of Southern California ♠University of California, Los Ange... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Kevin Frans | - | Offline RL Performance Bottlenecks | {'Is Value Learning Really the Main Bottleneck in Offline RL?': 'Title: Is Value Learning Really the Main Bottleneck in Offline RL?\\nAbstract \\nThe paper reports upon a series of depth interviews conducted with members of the \\nNASCAR culture of consumption. It was found that fans are committed to the drivers \\nthe... | {'Goal-Conditioned Data Augmentation for Offline Reinforcement Learning': 'Title: Goal-Conditioned Data Augmentation for Offline Reinforcement Learning\\nWhat is Essential for Unseen Goal Generalization ofOffline Goal-conditioned RL?Rui Yang 1 Yong Lin 1 Xiaoteng Ma 2 Hao Hu 2 Chongjie Zhang 3 Tong Zhang 1AbstractOffli... |
Vaskar Nath | - | Contrastive Goal-Conditioned Reward Learning | {'Learning Goal-Conditioned Representations for Language Reward Models': 'Title: Learning Goal-Conditioned Representations for Language Reward Models\\nAbstract\\nIt is of significance for an agent to learn a widely\\napplicable and general-purpose policy that can\\nachieve diverse goals including images and text\\ndes... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Jeff Da | - | Contrastive Goal-Conditioned Reward Learning | {'Learning Goal-Conditioned Representations for Language Reward Models': 'Title: Learning Goal-Conditioned Representations for Language Reward Models\\nAbstract\\nIt is of significance for an agent to learn a widely\\napplicable and general-purpose policy that can\\nachieve diverse goals including images and text\\ndes... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Hugh Zhang | - | Contrastive Goal-Conditioned Reward Learning | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Jennifer D'Souza | 0000-0002-6616-9509 | Ontology Matching with LLMs | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Large Language Models Struggle to Encode Medical Concepts - A Multilingual Benchmarking and Comparative Analysis': 'Title: Large Language Models Struggle to Encode Medical Concepts - A Multilingual Benchmarking and Comparative Analysis\\n1A Survey of GPT-3 Family Large LanguageModels Including ChatGPT and GPT-4Katika... |
Nathan Hu | - | null | {'Environmental Copper Sensor Based on Polyethylenimine-Functionalized Nanoporous Anodic Alumina Interferometers.': 'Title: Environmental Copper Sensor Based on Polyethylenimine-Functionalized Nanoporous Anodic Alumina Interferometers.\\nabstract. “Just Accepted” manuscripts have been fullypeer reviewed, but should not... | {'Electrokinetic Enhancement of Membrane Techniques for Efficient Nanoparticle Separation and Preconcentration.': 'Title: Electrokinetic Enhancement of Membrane Techniques for Efficient Nanoparticle Separation and Preconcentration.\\nElectrokinetic Trapping of Biomolecules:Novel Nanofluidic Devices for Proteomic Applic... |
Leyang Cui | 0000-0001-5072-884X | Machine-Generated Text Detection | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Tuney Zheng | - | Instruction Data Mining for LLM Reasoning | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Benchmarking Table Comprehension In The Wild': 'Title: Benchmarking Table Comprehension In The Wild\\nAbstract\\nZero-shot text classification (0SHOT-TC) is a\\nchallenging NLU problem to which little at-\\ntention has been paid by the research com-\\nmunity. 0SHOT-TC aims to associate an ap-\\npropriate label with a... |
Juhwan Choi | - | LLM-based Dataset Cleansing | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... | {'Learning Free Token Reduction for Multi-Modal LLM': 'Title: Learning Free Token Reduction for Multi-Modal LLM\\nHallucination Augmented Contrastive Learning for Multimodal Large LanguageModelChaoya Jiang1 Haiyang Xu2∗ Mengfan Dong1 Jiaxing Chen1 Wei Ye1∗Ming Yan2 Qinghao Ye2 Ji Zhang2 Fei Huang2 Shikun Zhang11Nationa... |
Bhavya Sukhija | 0000-0001-6238-9734 | Nonepisodic Optimistic Reinforcement Learning | {'Optimistic Active Exploration of Dynamical Systems': 'Title: Optimistic Active Exploration of Dynamical Systems\\nOptimistic Active Exploration of Dynamical SystemsBhavya Sukhija1 Lenart Treven1 Cansu Sancaktar2 Sebastian Blaes2Stelian Coros1 Andreas Krause1ETH Zürich 1 MPI for Intelligent Systems2{sukhijab,trevenl,s... | {'Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models': 'Title: Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models\\nABSTRACT\\nApproximations in computing model likelihoods with continuous normalizing\\nflows (CNFs) hinder the use of these models for importance sampling... |
Songhwai Oh | - | Adversarial Environment Design | {'Adversarial Environment Design via Regret-Guided Diffusion Models': 'Title: Adversarial Environment Design via Regret-Guided Diffusion Models\\nAbstract\\nTraining agents that are robust to environmental changes remains a significant\\nchallenge in deep reinforcement learning (RL). Unsupervised environment design\\n(... | {'Adversarial Environment Design via Regret-Guided Diffusion Models': 'Title: Adversarial Environment Design via Regret-Guided Diffusion Models\\nAbstract\\nTraining agents that are robust to environmental changes remains a significant\\nchallenge in deep reinforcement learning (RL). Unsupervised environment design\\n(... |
Markus Stocker | 0000-0001-5492-3212 | Scholarly Knowledge Graph | {'Managing Comprehensive Research Instrument Descriptions Within a Scholarly Knowledge Graph': 'Title: Managing Comprehensive Research Instrument Descriptions Within a Scholarly Knowledge Graph\\nabstract are my original work. Proper attribution \\nhas been given to all outside sources. I understand that I am solely re... | {'Resources for Graph Data and Knowledge': 'Title: Resources for Graph Data and Knowledge\\n City, University of London Institutional RepositoryCitation: Jimenez-Ruiz, E. ORCID: 0000-0002-9083-4599, Hassanzadeh, O., Efthymiou, V., Chen, J. and Srinivas, K. (2020). SemTab 2019: Resources to Benchmark Tabula... |
Shouhong Ding | - | Enhanced Deepfake Detection with Diffusion Models | {'Exploring Frequency Adversarial Attacks for Face Forgery Detection': 'Title: Exploring Frequency Adversarial Attacks for Face Forgery Detection\\n1Exploring Decision-based Black-box Attacks onFace Forgery DetectionZhaoyu Chen, Graduate Student Member, IEEE, Bo Li, Kaixun Jiang,Shuang Wu, Shouhong Ding, Wenqiang Zhang... | {'Hybrid Neural Network Models to Estimate Vital Signs from Facial Videos': 'Title: Hybrid Neural Network Models to Estimate Vital Signs from Facial Videos\\nAbstract: Generative artificial intelligence (AI) tools offer promise in facilitating code discovery, refinement, and large-scale analyses in qualitative coding. ... |
Zhiqing Sun | - | Easy-to-Hard Generalization in AI Alignment | {'Improving Reinforcement Learning from Human Feedback with Efficient Reward Model Ensemble': 'Title: Improving Reinforcement Learning from Human Feedback with Efficient Reward Model Ensemble\\nImproving the generalizability and robustness of large-scale traffic signal controlTianyu ShiDepartment of Civil Engineering, ... | {'Adversarial Environment Design via Regret-Guided Diffusion Models': 'Title: Adversarial Environment Design via Regret-Guided Diffusion Models\\nAbstract\\nTraining agents that are robust to environmental changes remains a significant\\nchallenge in deep reinforcement learning (RL). Unsupervised environment design\\n(... |
Longhui Yu | - | Easy-to-Hard Generalization in AI Alignment | {'Memory Replay with Data Compression for Continual Learning': 'Title: Memory Replay with Data Compression for Continual Learning\\nAbstract\\nDespite advances in deep learning, neural net-\\nworks can only learn multiple tasks when trained\\non them jointly. When tasks arrive sequentially,\\nthey lose performance on p... | {'Object Style Diffusion for Generalized Object Detection in Urban Scene': 'Title: Object Style Diffusion for Generalized Object Detection in Urban Scene\\nabstract style in the feature space. In the single-domain urban object detectiontask, where the style information is limited, we aim to extract the style informatio... |
Weiyang Liu | - | Easy-to-Hard Generalization in AI Alignment | {'Representational Alignment Supports Effective Machine Teaching': "Title: Representational Alignment Supports Effective Machine Teaching\\nRepresentational AlignmentSupports Effective Machine TeachingIlia Sucholutsky∗Princeton UniversityKatherine M. Collins∗University of CambridgeMaya MalaviyaStevens Institute of Tech... | {': Large': 'Title: : Large\\nAbstract. A rainbow subgraph of an edge-colored graph is a subgraph whose edges have\\ndistinct colors. The color degree of a vertex v is the number of different colors on edges\\nincident to v. We show that if n is large enough (namely, n ≥ 4.25k2), then each n-vertex\\ngraph G with minim... |
Yonghyeon Jo | - | Offline Reinforcement Learning | {'Adversarial Environment Design via Regret-Guided Diffusion Models': 'Title: Adversarial Environment Design via Regret-Guided Diffusion Models\\nAbstract\\nTraining agents that are robust to environmental changes remains a significant\\nchallenge in deep reinforcement learning (RL). Unsupervised environment design\\n(... | {'Adversarial Environment Design via Regret-Guided Diffusion Models': 'Title: Adversarial Environment Design via Regret-Guided Diffusion Models\\nAbstract\\nTraining agents that are robust to environmental changes remains a significant\\nchallenge in deep reinforcement learning (RL). Unsupervised environment design\\n(... |
Florent Teichteil-Konigsbuch | 0000-0001-6815-096X | Skill Diversity in Reinforcement Learning | {'Influence of modeling structure in probabilistic sequential decision problems': 'Title: Influence of modeling structure in probabilistic sequential decision problems\\nSequential Influence Diagrams: A Unified Asymmetry FrameworkFinn V. Jensen and Thomas D. NielsenDepartment of Computer ScienceAalborg UniversityFredri... | {'Adversarial Environment Design via Regret-Guided Diffusion Models': 'Title: Adversarial Environment Design via Regret-Guided Diffusion Models\\nAbstract\\nTraining agents that are robust to environmental changes remains a significant\\nchallenge in deep reinforcement learning (RL). Unsupervised environment design\\n(... |
Zhixun Li | 0000-0001-6750-9002 | Graph Training Debugging with Dynamic Soft-Pruning | {'Fairness without Demographics through Learning Graph of Gradients': 'Title: Fairness without Demographics through Learning Graph of Gradients\\nABSTRACT\\nMuch of the previous machine learning (ML) fairness literature\\nassumes that protected features such as race and sex are present\\nin the dataset, and relies upon... | {'Revisiting the Necessity of Graph Learning and Common Graph Benchmarks': 'Title: Revisiting the Necessity of Graph Learning and Common Graph Benchmarks\\nRevisiting the Message Passing inHeterophilous Graph Neural NetworksZhuonan Zheng1, Yuanchen Bei1, Sheng Zhou1∗, Yao Ma2,Ming Gu1, Hongjia Xu1, Chengyu Lai1, Jiawei... |
Guande He | - | Alignment of Language Models via Noise Contrastive Estimation | {'Noise Contrastive Alignment of Language Models with Explicit Rewards': 'Title: Noise Contrastive Alignment of Language Models with Explicit Rewards\\nabstract away imperceptible details and reduce computation, as thoselatent spaces still carry spatial details. On the other hand, it is a well-known fact that pretraine... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Lifan Yuan | - | Alignment of Language Models via Noise Contrastive Estimation | {'Noise Contrastive Alignment of Language Models with Explicit Rewards': 'Title: Noise Contrastive Alignment of Language Models with Explicit Rewards\\nabstract away imperceptible details and reduce computation, as thoselatent spaces still carry spatial details. On the other hand, it is a well-known fact that pretraine... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Edgar Dobriban | - | Dualized Alignment for LLMs | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Francesco Bacchiocchi | - | Online Bayesian Persuasion with Unknown Priors | {'Online Bayesian Persuasion Without a Clue': 'Title: Online Bayesian Persuasion Without a Clue\\nabstract ideas. (R)11 Extraversion I talk to a lot of different people at parties.12 Agreeableness I feel others’ emotions.13 Conscientiousness I like order.14 Neuroticism I get upset easily.15 Openness I gave difficulty u... | {'Learning in Repeated Multi-Unit Pay-As-Bid Auctions': 'Title: Learning in Repeated Multi-Unit Pay-As-Bid Auctions\\nabstract=678301Kira Goldner, Nicole Immorlica, and Brendan Lucier. 2019. Reducing Inefficiency in Carbon Auctions withImperfect Competition. https://doi.org/10.48550/ARXIV.1912.06428Negin Golrezaei, Pat... |
Nicola Gatti | - | Online Bayesian Persuasion with Unknown Priors | {'Online Bayesian Persuasion': 'Title: Online Bayesian Persuasion\\nAbstract\\nBayesian persuasion studies how an informed\\nsender should partially disclose information to in-\\nfluence the behavior of a self-interested receiver.\\nClassical models make the stringent assumption\\nthat the sender knows the receiver’s u... | {'Strategy-Proofness Made Simpler': 'Title: Strategy-Proofness Made Simpler\\nAbstract\\nEmpirical tests of direct revelation games of strategy-proof social choice functions\\nshow behavior in these games can approximate suboptimal mutual best responses. We\\ndetermine that this worst case scenario for the operation of... |
Emre Neftci | 0000-0002-0332-3273 | Deep Reinforcement Learning for Automatic Differentiation Optimization | {'Optimizing Automatic Differentiation with Deep Reinforcement Learning': 'Title: Optimizing Automatic Differentiation with Deep Reinforcement Learning\\nABSTRACT\\nGradient-based optimization is the foundation of deep learning and reinforcement\\nlearning, but is difficult to apply when the mechanism being optimized i... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Mohamad Yaser Jaradeh | 0000-0001-8777-2780 | Neuro-Symbolic Scholarly Search | {'Question Answering on Scholarly Knowledge Graphs': 'Title: Question Answering on Scholarly Knowledge Graphs\\nAbstract. Answering questions on scholarly knowledge comprising text\\nand other artifacts is a vital part of any research life cycle. Querying\\nscholarly knowledge and retrieving suitable answers is current... | {'Multimodal Multihop Source Retrieval for Web Question Answering': 'Title: Multimodal Multihop Source Retrieval for Web Question Answering\\nMPMQA: Multimodal Question Answering on Product ManualsLiang Zhang1, Anwen Hu1, Jing Zhang2, Shuo Hu2, Qin Jin1*1School of Information, Renmin University of China2Samsung Researc... |
Brian Mac Namee | 0000-0003-2518-0274 | Parallel Decoding for NER | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Low-Resource Chinese Named Entity Recognition via CNN-based Multitask Learning': 'Title: Low-Resource Chinese Named Entity Recognition via CNN-based Multitask Learning\\nabstract representation with two exceptions.The lowest layer takes input as a vector generated from the data sample. Thehighest layer usually output... |
Yaodong Yang | 0000-0001-8132-5613 | Preference-Adaptive AI Alignment | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Philhoon Oh | - | Key Value Fusion for Context-Order Invariance in LLMs | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Umut Simsekli | - | PDMP-Based Generative Models | {'Piecewise deterministic generative models': 'Title: Piecewise deterministic generative models\\nAbstract\\nThis paper introduces a novel mixture model-based approach for simultaneous clustering\\nand optimal segmentation of functional data which are curves presenting regime changes.\\nThe proposed model consists in a... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Dadi Guo | - | Privacy-Preserving Language Models | {'Multi-step Jailbreaking Privacy Attacks on ChatGPT': 'Title: Multi-step Jailbreaking Privacy Attacks on ChatGPT\\nMulti-step Jailbreaking Privacy Attacks on ChatGPTHaoran Li1, Dadi Guo2, Wei Fan1, Mingshi Xu1, Yangqiu Song11Dept. of CSE, Hong Kong University of Science and Technology2Center for Data Science, AAIS, Pe... | {'Safeguarding System Prompts for LLMs': 'Title: Safeguarding System Prompts for LLMs\\nCovert Malicious Finetuning: Challenges in Safeguarding LLM AdaptationDanny Halawi * 1 Alexander Wei * 1 Eric Wallace 1 Tony T. Wang 2 Nika Haghtalab † 1 Jacob Steinhardt † 1AbstractBlack-box finetuning is an emerging interface fora... |
Qi Hu | 0000-0002-4499-7805 | Privacy-Preserving Language Models | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Compare the Performance of Current Large Language Models in Terms of System Security': 'Title: Compare the Performance of Current Large Language Models in Terms of System Security\\nABSTRACT \\nCanada has observed wage differences between comparable immigrants and the Canadian-\\nborn across the labour market. Using ... |
Tanay Wakhare | - | Adversarial Prompt Engineering | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Rumen Dangovski | - | Quantum-informed Tensor Adaptation (QuanTA) | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Saleh Ashkboos | 0000-0001-6115-6779 | Quantized Rotational Inference | {'Towards End-to-end 4-Bit Inference on Generative Large Language Models': 'Title: Towards End-to-end 4-Bit Inference on Generative Large Language Models\\nQUIK: TOWARDS END-TO-END 4-BIT INFERENCEON GENERATIVE LARGE LANGUAGE MODELSSaleh Ashkboos * 1 Ilia Markov * 2 Elias Frantar 2 Tingxuan Zhong 3 Xingchen Wang 3 Jie R... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Maximilian L. Croci | - | Quantized Rotational Inference | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Pashmina Cameron | 0009-0009-0444-1755 | Quantized Rotational Inference | {'Development and evaluation of a multiscale keypoint detector based on complex wavelets': "Title: Development and evaluation of a multiscale keypoint detector based on complex wavelets\\nDevelopment and evaluation of a multiscalekeypoint detector based on complex waveletsPashmina BendaleChurchill College, CambridgeJan... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Jiaxuan You | - | Retrieval-Augmented Generation | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Mohammad Shoeybi | - | Retrieval-Augmented Generation | {'Retrieval meets Long Context Large Language Models': 'Title: Retrieval meets Long Context Large Language Models\\nPickens, J. & MacFarlane, A. (2006). Term Context Models for Information Retrieval. Paper presented at the 15th ACM international conference on Information and knowledge management, 05-11-2006 - 11-11-200... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Jonathan D. Chang | 0000-0002-3952-5475 | Reinforcement Learning Policy Optimization | {'Policy-Gradient Training of Language Models for Ranking': "Title: Policy-Gradient Training of Language Models for Ranking\\nPolicy-Gradient Training of Language Modelsfor RankingGe Gao Jonathan D. Chang Claire Cardie Kianté Brantley Thorsten JoachimsDepartment of Computer Science, Cornell University{ggao}@cs.cornell.... | {'Large Language Model Can Be a Foundation for Hidden Rationale-Based Retrieval': 'Title: Large Language Model Can Be a Foundation for Hidden Rationale-Based Retrieval\\nRethinking Interpretability in the Era of Large Language ModelsChandan Singh 1 Jeevana Priya Inala 1 Michel Galley 1 Rich Caruana 1 Jianfeng Gao 1Abst... |
Kianté Brantley | 0000-0002-8395-594X | Reinforcement Learning Policy Optimization | {'Policy-Gradient Training of Language Models for Ranking': "Title: Policy-Gradient Training of Language Models for Ranking\\nPolicy-Gradient Training of Language Modelsfor RankingGe Gao Jonathan D. Chang Claire Cardie Kianté Brantley Thorsten JoachimsDepartment of Computer Science, Cornell University{ggao}@cs.cornell.... | {'Inverse Reinforcement Learning from Demonstrations for LLM Alignment': 'Title: Inverse Reinforcement Learning from Demonstrations for LLM Alignment\\nInverse-RLignment: Inverse Reinforcement Learningfrom Demonstrations for LLM AlignmentHao Sun∗ Mihaela van der SchaarDepartment of Applied Mathematics and Theoretical P... |
Atticus Geiger | 0000-0002-9170-506X | Representation Finetuning (ReFT) | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Christopher Potts | 0000-0002-7978-6055 | Representation Finetuning (ReFT) | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'A Grounded Typology of Word Classes': 'Title: A Grounded Typology of Word Classes\\nUvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)UvA-DARE (Digital Academic Repository)Towards a mathematical model of word class clusteringsNordhoff, S.Publication date2008Published in... |
António Farinhas | - | Reranker-Enhanced Error Reduction in LLMs | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'The Impact of Model Scaling on Seen and Unseen Language Performance': 'Title: The Impact of Model Scaling on Seen and Unseen Language Performance\\nAbstract. First-person vision is gaining interest as it offers a unique\\nviewpoint on people’s interaction with objects, their attention, and even\\nintention. However, ... |
Haau-Sing Li | - | Reranker-Enhanced Error Reduction in LLMs | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Daniel Tang | 0000-0002-6377-0884 | Code Similarity Metrics | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Jiayi Ji | 0000-0002-9956-6308 | Spatial Awareness in 3D Segmentation | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Enhancing Fine-Grained Vision-Language Pretraining with Negative Augmented Samples': 'Title: Enhancing Fine-Grained Vision-Language Pretraining with Negative Augmented Samples\\nEnhancing Representation inRadiography-Reports Foundation Model: AGranular Alignment Algorithm Using MaskedContrastive LearningWeijian Huang... |
Daniil Dmitriev | - | Robust Mixture Estimation | {'Robust Mixture Learning when Outliers Overwhelm Small Groups': 'Title: Robust Mixture Learning when Outliers Overwhelm Small Groups\\nAbstract\\nIn this thesis, we study the problems of robust model selection and outlier detection\\nin linear regression. The \\nresults of data analysis based on linear regressions are... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Rares-Darius Buhai | 0000-0001-6667-0304 | Robust Mixture Estimation | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Fanny Yang | 0000-0003-0224-4079 | Robust Mixture Estimation | {'Semi-supervised novelty detection using ensembles with regularized disagreement': 'Title: Semi-supervised novelty detection using ensembles with regularized disagreement\\nA Survey on Semi-Supervised Learning for Delayed Partially\\nLabelled Data Streams\\nHEITOR MURILO GOMES, AI Institute, University of Waikato\\nMA... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Sahil Singla | 0000-0002-8800-6479 | Optimal Pricing with Limited Sampling | {'Online Combinatorial Allocations and Auctions with Few Samples': 'Title: Online Combinatorial Allocations and Auctions with Few Samples\\nAbstract\\nWe consider the online problem in which an intermediary trades identical items with a\\nsequence of n buyers and n sellers, each of unit demand. We assume that the value... | {'Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment': 'Title: Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment\\nABSTRACT\\nClassifier-Free Guidance (CFG) is a critical technique for enhancing the sample\\nquality of visual generative models. However, in autoregr... |
Justin C. Ondry | 0000-0001-9113-3420 | High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals | {} | {} |
Kailai Lin | 0000-0001-8509-5560 | High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals | {} | {} |
Benjamin F. Hammel | 0000-0002-9155-9875 | High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals | {} | {} |
Di Wang | 0000-0002-5841-5060 | High-Temperature Molten Salt Synthesis of III-V Semiconductor Nanocrystals | {} | {} |
Stella Chariton | 0000-0001-5522-0498 | High-Pressure Yttrium Hydrides Synthesis and Characterization | {'Superconductivity up to 243\\u2009K in the yttrium-hydrogen system under high pressure': 'Title: Superconductivity up to 243\\u2009K in the yttrium-hydrogen system under high pressure\\nSuperconductivity up to 243 K in yttrium hydrides under high pressure \\n \\nP. P. Kong1, V. S. Minkov1, M. A. Kuzovnikov1,5, S. P. ... | {'Local electronic properties of La3Ni2O7 under pressure': 'Title: Local electronic properties of La3Ni2O7 under pressure\\narXiv:2308.09044v1 [cond-mat.supr-con] 17 Aug 2023High-Tcsuperconductivity by mobilizing local spin singlets andpossible route to higher Tcin pressurized La3Ni2O7Qiong Qin1, 2 and Yi-feng Yang1,... |
Vitali Prakapenka | 0000-0001-9270-2330 | High-Pressure Yttrium Hydrides Synthesis and Characterization | {'Superconductivity up to 243 K in the yttrium-hydrogen system under high pressure': 'Title: Superconductivity up to 243 K in the yttrium-hydrogen system under high pressure\\nSuperconductivity up to 243 K in yttrium hydrides under high pressure \\n \\nP. P. Kong1, V. S. Minkov1, M. A. Kuzovnikov1,5, S. P. Besedin1, A.... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Tabish Iqbal | 0000-0002-2479-8729 | Enhanced 1-Alkene Production via Chimeric UndB-Catalase Biocatalyst | {'Functional production and biochemical investigation of an integral membrane enzyme for olefin biosynthesis': 'Title: Functional production and biochemical investigation of an integral membrane enzyme for olefin biosynthesis\\nAbstract ...................................................................................... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Subhashini Murugan | 0000-0002-4433-9191 | Enhanced 1-Alkene Production via Chimeric UndB-Catalase Biocatalyst | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'The Catalytic Mechanism of Key Enzymes Involved in the Synthesis of Ergothioneine in Lentinula edodes': 'Title: The Catalytic Mechanism of Key Enzymes Involved in the Synthesis of Ergothioneine in Lentinula edodes\\nSEPTEMBER27th-30th 2022CROWNE PLAZABelgrade, Serbia~ ~ ~ELECTRONICABSTRACT BOOK~ ~ ~ ~ ORGANIZING COM... |
Momoka Miyoshi | 0009-0009-0454-1887 | Self-Assembly of Titanate Nanosheets | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Jixiu Niu | 0009-0007-0041-9273 | Allochroic Copper Iodide LEDs | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
David Grass | 0000-0003-1781-0969 | Pump-Probe Microscopy for Pigment Identification in Art | {'Investigation of Artists Pigments with a Nonlinear Microscopy Technique': 'Title: Investigation of Artists Pigments with a Nonlinear Microscopy Technique\\nAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxii\\nCHAPTER\\n1. \\nIntroduction . . . . . . . . . . . . . . . . .... | {'On-Chip Polarization Light Microscopy': 'Title: On-Chip Polarization Light Microscopy\\nABSTRACT: Optical metasurfaces revolutionized the approach to moulding the propagation of \\nlight by enabling simultaneous control of the light phase, momentum, amplitude and polarization. \\nThus, instantaneous spectropolarimetr... |
Liting Tao | 0000-0001-7278-1115 | Vapor Deposition of Perovskite Heterojunctions for X-ray Detection | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Phase Modulation on Low‐Dimensional Perovskite for High Performance Blue Light Emitting Diodes': "Title: Phase Modulation on Low‐Dimensional Perovskite for High Performance Blue Light Emitting Diodes\\n 저 시-비 리- 경 지 2.0 한민 는 아래 조건 르는 경 에 한하여 게 l 저 물 복제, 포, 전송, 전시, 공연 송할 수 습니다. 다 과 같 조건 라야 합니다: l 하는, 저 물 나... |
Bowen Jin | - | Enhanced Perovskite Crystals for Radiation Detection | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Thermal Expansion Behavior of Halide Perovskite Single Crystals Across a Broad Temperature Range': 'Title: Thermal Expansion Behavior of Halide Perovskite Single Crystals Across a Broad Temperature Range\\nABSTRACT\\nHere I present a complete study of lattice dynamics of negative thermal ex-\\npansion (NTE) material ... |
Rui Feng | 0009-0008-4599-4447 | Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Yong-Qi Zhen | 0000-0002-1537-7666 | Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives': 'Title: Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives\\nDesign and synthesis of new1... |
Dongbo Wu | 0000-0003-4500-047X | Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Jin-Bu Xu | 0009-0005-4707-5186 | Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Lan Zhang | 0000-0002-7572-1035 | Regioselective Late-Stage Functionalization of Osimertinib via Palladium-Catalyzed C-H Arylation | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives': 'Title: Design, synthesis, anti-inflammatory evaluation, and molecular docking studies of novel quinazoline-4(3H)-one-2-carbothioamide derivatives\\nDesign and synthesis of new1... |
Bin Wu | 0000-0002-8152-4011 | Ion-Selective Membranes with Crown Ether-MOF Integration | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Xingya Li | 0000-0002-2824-6809 | Ion-Selective Membranes with Crown Ether-MOF Integration | {'Metal–Organic Framework–Based Ion–Selective Membranes': 'Title: Metal–Organic Framework–Based Ion–Selective Membranes\\nAbstract Inorganic contaminants, such as fluoride and arsenic are problematic inorganic contaminants due to major human health risk and relatively high levels of occurrence. Metal-organic frameworks... | {'MOF-derived Carbon-Based Materials for Energy-Related Applications.': "Title: MOF-derived Carbon-Based Materials for Energy-Related Applications.\\nYap, Min Hui and Fow, Kam Loon and Chen, George Zheng (2017) Synthesis and applications of MOF - derived porous nanostructures. Green Energy & Environment, 2 (3). pp. 218... |
Jindong Chen | 0000-0002-3940-9968 | Noncentrosymmetric Pnictide NLO Materials | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Local Strain Engineering of Two-Dimensional Transition Metal Dichalcogenides Towards Quantum Emitters': 'Title: Local Strain Engineering of Two-Dimensional Transition Metal Dichalcogenides Towards Quantum Emitters\\nQuantum-dot-like states in molybdenum disulfide nanostructures due to the interplay\\nof local surface... |
Huikang Jiang | - | Noncentrosymmetric Pnictide NLO Materials | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Ning Ye | 0000-0002-3679-4047 | Noncentrosymmetric Pnictide NLO Materials | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... |
Chao-Hui Zhang | 0009-0009-5002-1890 | Ultralight Integrated Anode for Lithium Metal Batteries | {'Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$': 'Title: Pressure-induced formation of cubic lutetium hydrides derived from trigonal LuH$_3$\\nPressure-induced formation of cubic lutetium hydrides derived from trigonal LuH 3\\nOwen Moulding,1Samuel Gallego-Parra,2Yingzheng Gao,1Pi... | {'Research Progress of Silicon/Carbon Anodes in Lithium-Ion Batteries': "Title: Research Progress of Silicon/Carbon Anodes in Lithium-Ion Batteries\\nAbstract. People's need for energy is growing as science and technology advance, and finding effective ways to store and use energy has become crucial in today's world. D... |
Bo-Zheng Liu | 0000-0002-4642-3688 | Ultralight Integrated Anode for Lithium Metal Batteries | {'Interfacial Characteristics of Na2FePO4F and Its Carbon Coated Material for Lithium/Sodium Hybrid Ion Battery': 'Title: Interfacial Characteristics of Na2FePO4F and Its Carbon Coated Material for Lithium/Sodium Hybrid Ion Battery\\nAbstract................................................................................. | {'Effect of CoO loading on electrochemical properties of activated carbon from sugarcane bagasse': 'Title: Effect of CoO loading on electrochemical properties of activated carbon from sugarcane bagasse\\nJournal of Analytical and Applied Pyrolysis 168 (2022) 105724Available online 23 September 20220165-2370/© 2022 The ... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.