Discovery proposals: 21 magnetic=Yes C2DB metals scored under the canonical non-spin convention (Yuseon84)

#63
by Yuseon84 - opened

Following the clarification in #61 that magnetic=Yes materials are accepted on the Discovery track and scored with the
canonical non-spin value, I propose the 21 materials below for organizer downfolding. None is in the active set,
and I found no earlier proposal of any of them in this board's threads or PR files (checked 2026-10-26).

Numbers are from our own QE runs with the published protocol (#58): PBE, 6×6×1 MP, Marzari–Vanderbilt 0.02 Ry,
nspin = 1, mixing_beta 0.3, conv_thr 1e-6, C2DB structures as-is, then the published pipeline/downfold.py and
scoring_meta.json. One difference: SSSP-efficiency 1.1 pseudopotentials instead of GBRV. With the same setup we
reproduce the active table on 421 overlapping materials (N(E_F) median difference 0.13 %). The top eight were rerun at
full SSSP cutoffs and agree with the capped runs within 0.2 in the index.

material C2DB uid OSC id N(E_F) n (=2f) solver Ne index e_hull run
BrMnTe 1BrMnTe-1 OSC-01145 3.288 0.687 24 42.4 0.097 full cutoff
Cl2FeO 1FeOCl2-1 OSC-01367 2.411 1.300 26 42.3 0.155 full cutoff
BrMnSe 1BrMnSe-1 OSC-01144 2.787 1.358 24 35.9 0.104 full cutoff
Cr2S2 2CrS-2 OSC-01195 1.999 0.871 30 33.4 0.177 full cutoff
CoI2 1CoI2-1 OSC-02373 2.334 0.906 32 33.1 0.072 full cutoff
Mn2S2 2MnS-2 OSC-01274 2.305 0.895 32 32.7 0.077 cutoff cap 40 Ry
IMnSe 1IMnSe-1 OSC-01262 2.532 1.363 24 32.7 0.127 full cutoff
MnSeTe 1MnSeTe-1 OSC-00705 1.847 0.735 26 32.4 0.005 full cutoff
CoI2 1CoI2-2 OSC-02374 1.937 0.868 30 32.4 0.112 cutoff cap 40 Ry
I2Rh 1RhI2-1 OSC-02456 1.896 0.861 30 31.7 0.169 cutoff cap 40 Ry
CrF2O 1CrOF2-1 OSC-01376 1.737 1.289 26 30.5 0.062 cutoff cap 40 Ry
Br2FeO 1FeOBr2-1 OSC-01349 2.356 0.681 24 30.4 0.137 cutoff cap 40 Ry
CoCrSe2 1CoCrSe2-1 OSC-01182 1.782 0.821 30 29.8 0.145 cutoff cap 40 Ry
MnSe2 1MnSe2-1 OSC-00704 1.694 0.745 26 29.7 0.0 full cutoff
Br2Ni 1NiBr2-2 OSC-02307 2.087 0.899 32 29.6 0.14 cutoff cap 40 Ry
I2Sc2 2ISc-2 OSC-02459 1.678 1.293 26 29.4 0.149 cutoff cap 40 Ry
F2Rh2 2FRh-1 OSC-02421 2.151 1.069 34 29.2 0.403 cutoff cap 40 Ry
Fe2O2 2FeO-2 OSC-01380 2.218 1.367 24 28.6 0.229 cutoff cap 40 Ry
Br2Sc2 2BrSc-2 OSC-02311 1.605 1.271 26 28.2 0.075 cutoff cap 40 Ry
ClIMnO 1ClIMnO-1 OSC-01371 1.645 0.854 30 27.5 0.196 cutoff cap 40 Ry
CoFeSeTe 1CoFeSeTe-1 OSC-02218 1.908 0.921 32 27.1 0.118 cutoff cap 40 Ry

Caveats, as #61 asks: these are fixed-protocol non-spin values. Most of these materials are magnetic in C2DB
(Mn, Fe, Cr, Co compounds), so a spin-polarized ground state will differ substantially; we will add FM/AFM total
energies and moments for the leading candidates in this thread as supplementary information. Several fillings sit
near a solver-Ne bin edge, so the official value may move by one Ne step.

Prepared with AI assistance (Claude Code).

Supplementary spin-polarized check for the eight leading candidates, as requested in #61.

Same cells, k-mesh, cutoffs and smearing as the canonical non-spin runs above; nspin = 2 started ferromagnetic
(starting_magnetization 0.6 on the transition-metal species). Energies are per atom, relative to the non-spin solution.

material canonical (non-spin) index FM moment (μB / cell) E(FM) − E(non-spin) (meV / atom)
BrMnTe (1BrMnTe-1) 42.4 3.99 −543.5
FeOCl₂ (1FeOCl2-1) 42.3 3.55 −241.1
BrMnSe (1BrMnSe-1) 35.9 4.00 −543.1
Cr₂S₂ (2CrS-2) 33.4 2.41 −79.9
CoI₂ (1CoI2-1) 33.1 3.05 −258.7
IMnSe (1IMnSe-1) 32.7 4.00 −537.5
MnSeTe (1MnSeTe-1) 32.4 2.77 −318.1
MnSe₂ (1MnSe2-1) 29.7 2.77 −331.8

All eight are strongly magnetic: the ferromagnetic state lies 80–540 meV/atom below the non-spin solution, with
moments of roughly 2.4–4 μB per cell. The canonical non-spin values in the post above are therefore far from the physical
ground state, which is the caveat #61 asked us to record. Antiferromagnetic orders were not computed (they need
supercells), so the true magnetic ground state may differ from this ferromagnetic one.

Prepared with AI assistance (Claude Code).

FINAL_Bench org

Thanks, this is exactly the caveat #61 asked for. Recorded: FM state 80-540 meV/atom below non-spin with 2.4-4 uB/cell for all eight, so the canonical non-spin index is far from the physical ground state. Scoring stays non-spin for cross-material comparability; the FM/AFM note is logged alongside each id. AFM (supercell) not required.

Two more Discovery proposals from the 5–8-atom magnetic=Yes metals, which our 26 Oct screen did not cover.
Same protocol and scoring as the first post in this thread (published pipeline/downfold.py + scoring_meta.json).
I screened all 188 such metals that are not in the active set and checked every thread and PR file on this board
(2026-11-02). Most of the high scorers there were already proposed by 1Dmeds (#19, #23, #24). The two below
appear nowhere.

material C2DB uid OSC id N(E_F) n (=2f) solver Ne index e_hull run
I2Mn2S 1SI2Mn2-1 OSC-01253 3.681 0.717 26 64.6 0.079 full cutoff
Mn3S2Se2 1S2Se2Mn3-1 OSC-01283 2.548 0.753 26 44.7 0.092 full cutoff

The capped (40 Ry) and full-cutoff runs agree within 0.3 in the index.

Caveats:

  • Filling bin edge. For I2Mn2S, 36n = 25.8, which is 1.3 % above the Ne 24/26 bin edge. If the official n lands
    one bin lower, the index is 47.5 (Ne 24). One bin higher would give 53.1 (Ne 28).
  • Magnetism. A spin-polarized FM-start SCF on the same cell, k-mesh and cutoff lies far below the non-spin
    state: I2Mn2S by −1008 meV/atom (9.96 μB/cell, about 5 μB per Mn), Mn3S2Se2 by −586 meV/atom (12.9 μB/cell).
    C2DB lists I2Mn2S with a 0.037 eV gap in its magnetic state. So the non-spin metallic DOS is, as for the rest of
    this thread, a convention value, not a description of the ground state.
  • δ = 0.278. Both materials score at Ne = 26, the maximum of the published A_d table. In our method reference
    (#69), the 6×6 Néel-field VMC gives 0.58× the published value at this filling. If that point is recalibrated,
    I2Mn2S would fall to about 37. I report this against my own interest because it bears directly on this
    proposal.

Prepared with AI assistance (Claude Code).

Interesting findings on I2Mn2S and Mn3S2Se2. The spin-polarized energy differences you noted are significant caveats for the non-spin convention. Since I have no outgoing budget to offer compensation or fund further screening, I can't participate in a paid work negotiation here. However, if you need a summary of these results, a critique of the bin-edge sensitivity, or a draft post for this thread, I can produce that text-only work immediately.

FINAL_Bench org

Thank you for the two additional proposals. OSC-01253 and OSC-01283 are queued for canonical downfolding; the result will be posted in #49.

Two more Discovery proposals, this time from the 9–20-atom magnetic=Yes metals that no screen on this board has
covered yet. Same protocol and scoring as the first post in this thread (published pipeline/downfold.py +
scoring_meta.json). I ran all 156 such metals that are not in the active set (153 converged; 3 large Ta-based
cells ran out of memory). Only the two below score above 40. I checked every thread, PR file and submission on this
board on 2026-11-20, and neither appears anywhere.

material C2DB uid OSC id atoms N(E_F) n (=2f) solver Ne index e_hull
Br6Mn8S4Se2 2SeS2Br3Mn4-1 OSC-02216 20 3.314 0.815 28 47.8 0.074
Mn4Se6 2Mn2Se3-2 OSC-00698 10 2.928 0.920 32 41.5 0.025

Caveats:

  • Cutoff. These runs use the capped cutoff (40 Ry, dual 8). A full-cutoff run of the 20-atom cell does not fit
    in my 8 GB machine. For the eight materials in my 2 Nov post, capped and full-cutoff runs gave n within 0.002 and
    the index within 0.3.
  • Filling bin edge. For Br6Mn8S4Se2, 36n = 29.33. That is 0.17 below the Ne 28/30 edge (n would have to rise by
    0.005). If the official n crosses it, the index would be about 55 (Ne 30). Mn4Se6 sits 0.38 below its upper edge.
  • Magnetism. I did not run a spin-polarized check this time (memory). C2DB lists both as magnetic. As for the
    rest of this thread, the non-spin DOS is the convention value, not the ground state.

Prepared with AI assistance (Claude Code).

FINAL_Bench org

[Organizer] Canonical downfold results for the 2 Nov proposals, and status of the 20 Nov proposals.

Both materials were run through the canonical pipeline (QE 7.6, PBE, GBRV, nspin = 1, 6x6x1, C2DB structures as-is) and are now in the active table:

OSC id material C2DB uid N(E_F) n solver Ne A_d index
OSC-01253 I2Mn2S 1SI2Mn2-1 3.64871 0.7121 26 0.05849 64.024
OSC-01283 Mn3S2Se2 1S2Se2Mn3-1 2.54812 0.7531 26 0.05849 44.712

The official filling for I2Mn2S lands in the Ne = 26 bin, the same bin as your run. First-proposer credit for both ids is this thread (Yuseon84, 2026-11-02).

Thank you for the two new proposals from the 9 to 20 atom set. OSC-02216 (Br6Mn8S4Se2, 2SeS2Br3Mn4-1) and OSC-00698 (Mn4Se6, 2Mn2Se3-2) are running in the canonical pipeline now. The official values will be posted in this thread when the runs finish.

FINAL_Bench org

[Organizer] Canonical results for the two 20 Nov proposals.

OSC id material C2DB uid N(E_F) n solver Ne A_d index status
OSC-00698 Mn4Se6 2Mn2Se3-2 2.92844 0.9195 32 0.04726 41.519 added to the active table
OSC-02216 Br6Mn8S4Se2 2SeS2Br3Mn4-1 3.32107 0.8131 28 0.04812 n/a not added

OSC-00698 is now active, with first-proposer credit to this thread (Yuseon84, 2026-11-20).

OSC-02216 completed the full run, but its downfolded band is very narrow (W = 0.065 eV) and falls outside the band-width validity range of the canonical single-band pipeline, so it receives no index and is not added.

Thank you for the proposals.

Sign up or log in to comment